API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Aug 18th, 2019
796
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 417.85 KB | None | 0 0
raw download clone embed print report
  1. WARNING: Logging before InitGoogleLogging() is written to STDERR
  2. I0818 13:35:06.854073 18451 parallel.cpp:47] P2PManager::Init @ ntu329-ESC8000-G3-Series-Invalid-entry-length-16-Fixed-up-to-11
  3. I0818 13:35:07.238003 18451 caffe.cpp:709] This is NVCaffe 0.17.2 started at Sun Aug 18 13:35:06 2019
  4. I0818 13:35:07.238201 18451 caffe.cpp:711] CuDNN version: 7103
  5. I0818 13:35:07.238207 18451 caffe.cpp:712] CuBLAS version: 9000
  6. I0818 13:35:07.238209 18451 caffe.cpp:713] CUDA version: 9000
  7. I0818 13:35:07.238212 18451 caffe.cpp:714] CUDA driver version: 9000
  8. I0818 13:35:07.238217 18451 caffe.cpp:715] Arguments:
  9. [0]: ./build/tools/caffe
  10. [1]: train
  11. [2]: --solver=examples/cifar10/cifar10_vgg7_BN_solver.prototxt
  12. [3]: -gpu
  13. [4]: all
  14. I0818 13:35:07.240495 18451 caffe.cpp:220] Using GPUs 0, 1, 2, 3, 4, 5, 6, 7
  15. I0818 13:35:07.270864 18451 gpu_memory.cpp:105] GPUMemory::Manager initialized
  16. I0818 13:35:07.272593 18451 gpu_memory.cpp:107] Total memory: 11714691072, Free: 11239358464, dev_info[0]: total=11714691072 free=11239358464
  17. I0818 13:35:07.274250 18451 gpu_memory.cpp:107] Total memory: 11715084288, Free: 11239358464, dev_info[1]: total=11714691072 free=11239358464
  18. I0818 13:35:07.277063 18451 gpu_memory.cpp:107] Total memory: 11715084288, Free: 11239358464, dev_info[2]: total=11714691072 free=11239358464
  19. I0818 13:35:07.278730 18451 gpu_memory.cpp:107] Total memory: 11715084288, Free: 11239358464, dev_info[3]: total=11714691072 free=11239358464
  20. I0818 13:35:07.280411 18451 gpu_memory.cpp:107] Total memory: 11715084288, Free: 11239358464, dev_info[4]: total=11714691072 free=11239358464
  21. I0818 13:35:07.282086 18451 gpu_memory.cpp:107] Total memory: 11715084288, Free: 11239358464, dev_info[5]: total=11714691072 free=11239358464
  22. I0818 13:35:07.283978 18451 gpu_memory.cpp:107] Total memory: 11715084288, Free: 11239358464, dev_info[6]: total=11714691072 free=11239358464
  23. I0818 13:35:07.285643 18451 gpu_memory.cpp:107] Total memory: 11715084288, Free: 11239358464, dev_info[7]: total=11714691072 free=11239358464
  24. I0818 13:35:07.286790 18451 caffe.cpp:227] GPU 0: GeForce GTX 1080 Ti
  25. I0818 13:35:07.288208 18451 caffe.cpp:227] GPU 1: GeForce GTX 1080 Ti
  26. I0818 13:35:07.289592 18451 caffe.cpp:227] GPU 2: GeForce GTX 1080 Ti
  27. I0818 13:35:07.290987 18451 caffe.cpp:227] GPU 3: GeForce GTX 1080 Ti
  28. I0818 13:35:07.292412 18451 caffe.cpp:227] GPU 4: GeForce GTX 1080 Ti
  29. I0818 13:35:07.293853 18451 caffe.cpp:227] GPU 5: GeForce GTX 1080 Ti
  30. I0818 13:35:07.295294 18451 caffe.cpp:227] GPU 6: GeForce GTX 1080 Ti
  31. I0818 13:35:07.297849 18451 caffe.cpp:227] GPU 7: GeForce GTX 1080 Ti
  32. I0818 13:35:07.297998 18451 solver.cpp:40] Solver data type: FLOAT16
  33. I0818 13:35:07.323930 18451 solver.cpp:43] Initializing solver from parameters:
  34. test_iter: 100
  35. test_interval: 1000
  36. base_lr: 0.001
  37. display: 200
  38. max_iter: 50000
  39. lr_policy: "multistep"
  40. gamma: 0.1
  41. momentum: 0.9
  42. weight_decay: 0.004
  43. solver_mode: GPU
  44. device_id: 0
  45. net: "examples/cifar10/cifar10_vgg7_BN.prototxt"
  46. train_state {
  47. level: 0
  48. stage: ""
  49. }
  50. stepvalue: 30000
  51. stepvalue: 40000
  52. solver_data_type: FLOAT16
  53. I0818 13:35:07.324601 18451 solver.cpp:85] Creating training net from net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  54. I0818 13:35:07.325131 18451 net.cpp:459] The NetState phase (0) differed from the phase (1) specified by a rule in layer cifar
  55. I0818 13:35:07.325170 18451 net.cpp:459] The NetState phase (0) differed from the phase (1) specified by a rule in layer accuracy
  56. W0818 13:35:07.325253 18451 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  57. I0818 13:35:07.325577 18451 net.cpp:83] Initializing net from parameters:
  58. name: "CIFAR10_full_vgg7_BN"
  59. state {
  60. phase: TRAIN
  61. level: 0
  62. stage: ""
  63. }
  64. default_forward_type: FLOAT16
  65. default_backward_type: FLOAT16
  66. default_forward_math: FLOAT16
  67. default_backward_math: FLOAT16
  68. global_grad_scale: 2000
  69. layer {
  70. name: "cifar"
  71. type: "Data"
  72. top: "data"
  73. top: "label"
  74. include {
  75. phase: TRAIN
  76. }
  77. transform_param {
  78. scale: 0.00390625
  79. mirror: true
  80. crop_size: 28
  81. mean_file: "examples/cifar10/mean.binaryproto"
  82. }
  83. data_param {
  84. source: "examples/cifar10/cifar10_train_lmdb"
  85. batch_size: 13
  86. backend: LMDB
  87. }
  88. }
  89. layer {
  90. name: "conv1_1"
  91. type: "Convolution"
  92. bottom: "data"
  93. top: "conv1_1"
  94. param {
  95. lr_mult: 1
  96. }
  97. param {
  98. lr_mult: 2
  99. }
  100. convolution_param {
  101. num_output: 128
  102. pad: 1
  103. kernel_size: 3
  104. stride: 1
  105. weight_filler {
  106. type: "gspt_xavier"
  107. }
  108. bias_filler {
  109. type: "gspt_constant"
  110. }
  111. }
  112. }
  113. layer {
  114. name: "conv1_1_bn"
  115. type: "BatchNorm"
  116. bottom: "conv1_1"
  117. top: "conv1_1"
  118. batch_norm_param {
  119. use_global_stats: false
  120. }
  121. }
  122. layer {
  123. name: "relu1_1"
  124. type: "ReLU"
  125. bottom: "conv1_1"
  126. top: "conv1_1"
  127. }
  128. layer {
  129. name: "conv1_2"
  130. type: "Convolution"
  131. bottom: "conv1_1"
  132. top: "conv1_2"
  133. param {
  134. lr_mult: 1
  135. }
  136. param {
  137. lr_mult: 2
  138. }
  139. convolution_param {
  140. num_output: 128
  141. pad: 1
  142. kernel_size: 3
  143. stride: 1
  144. weight_filler {
  145. type: "gspt_xavier"
  146. }
  147. bias_filler {
  148. type: "gspt_constant"
  149. }
  150. }
  151. }
  152. layer {
  153. name: "conv1_2_bn"
  154. type: "BatchNorm"
  155. bottom: "conv1_2"
  156. top: "conv1_2"
  157. batch_norm_param {
  158. use_global_stats: false
  159. }
  160. }
  161. layer {
  162. name: "relu1_2"
  163. type: "ReLU"
  164. bottom: "conv1_2"
  165. top: "conv1_2"
  166. }
  167. layer {
  168. name: "pool1"
  169. type: "Pooling"
  170. bottom: "conv1_2"
  171. top: "pool1"
  172. pooling_param {
  173. pool: MAX
  174. kernel_size: 2
  175. stride: 2
  176. }
  177. }
  178. layer {
  179. name: "conv2_1"
  180. type: "Convolution"
  181. bottom: "pool1"
  182. top: "conv2_1"
  183. param {
  184. lr_mult: 1
  185. }
  186. param {
  187. lr_mult: 2
  188. }
  189. convolution_param {
  190. num_output: 256
  191. pad: 1
  192. kernel_size: 3
  193. stride: 1
  194. weight_filler {
  195. type: "gspt_xavier"
  196. }
  197. bias_filler {
  198. type: "gspt_constant"
  199. }
  200. }
  201. }
  202. layer {
  203. name: "conv2_1_bn"
  204. type: "BatchNorm"
  205. bottom: "conv2_1"
  206. top: "conv2_1"
  207. batch_norm_param {
  208. use_global_stats: false
  209. }
  210. }
  211. layer {
  212. name: "relu2_1"
  213. type: "ReLU"
  214. bottom: "conv2_1"
  215. top: "conv2_1"
  216. }
  217. layer {
  218. name: "conv2_2"
  219. type: "Convolution"
  220. bottom: "conv2_1"
  221. top: "conv2_2"
  222. param {
  223. lr_mult: 1
  224. }
  225. param {
  226. lr_mult: 2
  227. }
  228. convolution_param {
  229. num_output: 256
  230. pad: 1
  231. kernel_size: 3
  232. stride: 1
  233. weight_filler {
  234. type: "gspt_xavier"
  235. }
  236. bias_filler {
  237. type: "gspt_constant"
  238. }
  239. }
  240. }
  241. layer {
  242. name: "conv2_2_bn"
  243. type: "BatchNorm"
  244. bottom: "conv2_2"
  245. top: "conv2_2"
  246. batch_norm_param {
  247. use_global_stats: false
  248. }
  249. }
  250. layer {
  251. name: "relu2_2"
  252. type: "ReLU"
  253. bottom: "conv2_2"
  254. top: "conv2_2"
  255. }
  256. layer {
  257. name: "pool2"
  258. type: "Pooling"
  259. bottom: "conv2_2"
  260. top: "pool2"
  261. pooling_param {
  262. pool: MAX
  263. kernel_size: 2
  264. stride: 2
  265. }
  266. }
  267. layer {
  268. name: "conv3_1"
  269. type: "Convolution"
  270. bottom: "pool2"
  271. top: "conv3_1"
  272. param {
  273. lr_mult: 1
  274. }
  275. param {
  276. lr_mult: 2
  277. }
  278. convolution_param {
  279. num_output: 512
  280. pad: 1
  281. kernel_size: 3
  282. stride: 1
  283. weight_filler {
  284. type: "gspt_xavier"
  285. }
  286. bias_filler {
  287. type: "gspt_constant"
  288. }
  289. }
  290. }
  291. layer {
  292. name: "conv3_1_bn"
  293. type: "BatchNorm"
  294. bottom: "conv3_1"
  295. top: "conv3_1"
  296. batch_norm_param {
  297. use_global_stats: false
  298. }
  299. }
  300. layer {
  301. name: "relu3_1"
  302. type: "ReLU"
  303. bottom: "conv3_1"
  304. top: "conv3_1"
  305. }
  306. layer {
  307. name: "conv3_2"
  308. type: "Convolution"
  309. bottom: "conv3_1"
  310. top: "conv3_2"
  311. param {
  312. lr_mult: 1
  313. }
  314. param {
  315. lr_mult: 2
  316. }
  317. convolution_param {
  318. num_output: 512
  319. pad: 1
  320. kernel_size: 3
  321. stride: 1
  322. weight_filler {
  323. type: "gspt_xavier"
  324. }
  325. bias_filler {
  326. type: "gspt_constant"
  327. }
  328. }
  329. }
  330. layer {
  331. name: "conv3_2_bn"
  332. type: "BatchNorm"
  333. bottom: "conv3_2"
  334. top: "conv3_2"
  335. batch_norm_param {
  336. use_global_stats: false
  337. }
  338. }
  339. layer {
  340. name: "relu3_2"
  341. type: "ReLU"
  342. bottom: "conv3_2"
  343. top: "conv3_2"
  344. }
  345. layer {
  346. name: "pool3"
  347. type: "Pooling"
  348. bottom: "conv3_2"
  349. top: "pool3"
  350. pooling_param {
  351. pool: MAX
  352. kernel_size: 2
  353. stride: 2
  354. }
  355. }
  356. layer {
  357. name: "ip1"
  358. type: "InnerProduct"
  359. bottom: "pool3"
  360. top: "ip1"
  361. param {
  362. lr_mult: 1
  363. decay_mult: 0
  364. }
  365. param {
  366. lr_mult: 2
  367. decay_mult: 0
  368. }
  369. inner_product_param {
  370. num_output: 1024
  371. weight_filler {
  372. type: "gspt_xavier"
  373. }
  374. bias_filler {
  375. type: "gspt_constant"
  376. }
  377. }
  378. }
  379. layer {
  380. name: "relu4"
  381. type: "ReLU"
  382. bottom: "ip1"
  383. top: "ip1"
  384. }
  385. layer {
  386. name: "drop7"
  387. type: "Dropout"
  388. bottom: "ip1"
  389. top: "ip1"
  390. dropout_param {
  391. dropout_ratio: 0.5
  392. }
  393. }
  394. layer {
  395. name: "ip2"
  396. type: "InnerProduct"
  397. bottom: "ip1"
  398. top: "ip2"
  399. param {
  400. lr_mult: 1
  401. decay_mult: 0
  402. }
  403. param {
  404. lr_mult: 2
  405. decay_mult: 0
  406. }
  407. inner_product_param {
  408. num_output: 10
  409. weight_filler {
  410. type: "gspt_xavier"
  411. }
  412. bias_filler {
  413. type: "gspt_constant"
  414. }
  415. }
  416. }
  417. layer {
  418. name: "loss"
  419. type: "SoftmaxWithLoss"
  420. bottom: "ip2"
  421. bottom: "label"
  422. top: "loss"
  423. }
  424. I0818 13:35:07.325773 18451 layer_factory.hpp:172] Creating layer 'cifar' of type 'Data'
  425. I0818 13:35:07.325788 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT Fmath:FLOAT16 Bmath:FLOAT16
  426. I0818 13:35:07.325945 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 0
  427. I0818 13:35:07.326050 18451 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  428. I0818 13:35:07.329493 18451 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  429. I0818 13:35:07.329501 18498 blocking_queue.cpp:40] Data layer prefetch queue empty
  430. I0818 13:35:07.329682 18451 net.cpp:202] Created Layer cifar (0)
  431. I0818 13:35:07.329695 18451 net.cpp:544] cifar -> data
  432. I0818 13:35:07.329716 18451 net.cpp:544] cifar -> label
  433. I0818 13:35:07.329741 18451 data_reader.cpp:59] Sample Data Reader threads: 1, out queues: 1, depth: 13
  434. I0818 13:35:07.329766 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 0
  435. I0818 13:35:07.331662 18499 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  436. I0818 13:35:07.331912 18451 data_layer.cpp:200] [n0.d0.r0] Output data size: 13, 3, 28, 28
  437. I0818 13:35:07.331935 18451 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  438. I0818 13:35:07.332056 18451 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  439. I0818 13:35:07.332166 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 0
  440. I0818 13:35:07.332631 18451 net.cpp:262] Setting up cifar
  441. I0818 13:35:07.332665 18451 net.cpp:269] TRAIN Top shape for layer 0 'cifar' 13 3 28 28 (30576)
  442. I0818 13:35:07.332674 18451 net.cpp:269] TRAIN Top shape for layer 0 'cifar' 13 (13)
  443. I0818 13:35:07.334326 18451 layer_factory.hpp:172] Creating layer 'conv1_1' of type 'Convolution'
  444. I0818 13:35:07.334334 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  445. I0818 13:35:07.334405 18451 net.cpp:202] Created Layer conv1_1 (1)
  446. I0818 13:35:07.334415 18451 net.cpp:574] conv1_1 <- data
  447. I0818 13:35:07.334429 18451 net.cpp:544] conv1_1 -> conv1_1
  448. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  449. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  450. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  451. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  452. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  453. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  454. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  455. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  456. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  457. in math_function.cpp, line 730, tmp_value = 1048576
  458. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  459. definition of type is half.
  460. gspt r[i] = 0.0625 (Q_frac=24
  461. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  462. In blob.cpp, blob.Q_frac= 24
  463. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  464. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  465. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  466. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  467. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  468. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  469. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  470. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  471. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  472. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  473. in math_function.cpp, line 1227, tmp_value = 0
  474. definition of type is half [FILL ZERO].
  475. gspt r[i] = 0 (Q_frac=22
  476. In blob.cpp, blob.Q_frac= 22
  477. I0818 13:35:08.120280 18451 net.cpp:262] Setting up conv1_1
  478. I0818 13:35:08.120345 18451 net.cpp:269] TRAIN Top shape for layer 1 'conv1_1' 13 128 28 28 (1304576)
  479. I0818 13:35:08.120410 18451 layer_factory.hpp:172] Creating layer 'conv1_1_bn' of type 'BatchNorm'
  480. I0818 13:35:08.120434 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  481. I0818 13:35:08.120467 18451 net.cpp:202] Created Layer conv1_1_bn (2)
  482. I0818 13:35:08.120476 18451 net.cpp:574] conv1_1_bn <- conv1_1
  483. I0818 13:35:08.120486 18451 net.cpp:529] conv1_1_bn -> conv1_1 (in-place)
  484. I0818 13:35:08.121035 18451 net.cpp:262] Setting up conv1_1_bn
  485. I0818 13:35:08.121048 18451 net.cpp:269] TRAIN Top shape for layer 2 'conv1_1_bn' 13 128 28 28 (1304576)
  486. I0818 13:35:08.121068 18451 layer_factory.hpp:172] Creating layer 'relu1_1' of type 'ReLU'
  487. I0818 13:35:08.121083 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  488. I0818 13:35:08.121105 18451 net.cpp:202] Created Layer relu1_1 (3)
  489. I0818 13:35:08.121111 18451 net.cpp:574] relu1_1 <- conv1_1
  490. I0818 13:35:08.121117 18451 net.cpp:529] relu1_1 -> conv1_1 (in-place)
  491. I0818 13:35:08.121140 18451 net.cpp:262] Setting up relu1_1
  492. I0818 13:35:08.121147 18451 net.cpp:269] TRAIN Top shape for layer 3 'relu1_1' 13 128 28 28 (1304576)
  493. I0818 13:35:08.121152 18451 layer_factory.hpp:172] Creating layer 'conv1_2' of type 'Convolution'
  494. I0818 13:35:08.121158 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  495. I0818 13:35:08.121213 18451 net.cpp:202] Created Layer conv1_2 (4)
  496. I0818 13:35:08.121220 18451 net.cpp:574] conv1_2 <- conv1_1
  497. I0818 13:35:08.121225 18451 net.cpp:544] conv1_2 -> conv1_2
  498. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  499. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  500. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  501. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  502. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  503. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  504. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  505. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  506. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  507. in math_function.cpp, line 730, tmp_value = 1048576
  508. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  509. definition of type is half.
  510. gspt r[i] = 0.0078125 (Q_frac=27
  511. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  512. In blob.cpp, blob.Q_frac= 27
  513. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  514. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  515. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  516. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  517. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  518. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  519. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  520. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  521. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  522. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  523. in math_function.cpp, line 1227, tmp_value = 0
  524. definition of type is half [FILL ZERO].
  525. gspt r[i] = 0 (Q_frac=22
  526. In blob.cpp, blob.Q_frac= 22
  527. I0818 13:35:08.170953 18451 net.cpp:262] Setting up conv1_2
  528. I0818 13:35:08.170990 18451 net.cpp:269] TRAIN Top shape for layer 4 'conv1_2' 13 128 28 28 (1304576)
  529. I0818 13:35:08.171005 18451 layer_factory.hpp:172] Creating layer 'conv1_2_bn' of type 'BatchNorm'
  530. I0818 13:35:08.171012 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  531. I0818 13:35:08.171034 18451 net.cpp:202] Created Layer conv1_2_bn (5)
  532. I0818 13:35:08.171041 18451 net.cpp:574] conv1_2_bn <- conv1_2
  533. I0818 13:35:08.171048 18451 net.cpp:529] conv1_2_bn -> conv1_2 (in-place)
  534. I0818 13:35:08.171542 18451 net.cpp:262] Setting up conv1_2_bn
  535. I0818 13:35:08.171553 18451 net.cpp:269] TRAIN Top shape for layer 5 'conv1_2_bn' 13 128 28 28 (1304576)
  536. I0818 13:35:08.171572 18451 layer_factory.hpp:172] Creating layer 'relu1_2' of type 'ReLU'
  537. I0818 13:35:08.171578 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  538. I0818 13:35:08.171588 18451 net.cpp:202] Created Layer relu1_2 (6)
  539. I0818 13:35:08.171593 18451 net.cpp:574] relu1_2 <- conv1_2
  540. I0818 13:35:08.171598 18451 net.cpp:529] relu1_2 -> conv1_2 (in-place)
  541. I0818 13:35:08.171608 18451 net.cpp:262] Setting up relu1_2
  542. I0818 13:35:08.171615 18451 net.cpp:269] TRAIN Top shape for layer 6 'relu1_2' 13 128 28 28 (1304576)
  543. I0818 13:35:08.171622 18451 layer_factory.hpp:172] Creating layer 'pool1' of type 'Pooling'
  544. I0818 13:35:08.171625 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  545. I0818 13:35:08.171643 18451 net.cpp:202] Created Layer pool1 (7)
  546. I0818 13:35:08.171649 18451 net.cpp:574] pool1 <- conv1_2
  547. I0818 13:35:08.171655 18451 net.cpp:544] pool1 -> pool1
  548. I0818 13:35:08.171773 18451 net.cpp:262] Setting up pool1
  549. I0818 13:35:08.171784 18451 net.cpp:269] TRAIN Top shape for layer 7 'pool1' 13 128 14 14 (326144)
  550. I0818 13:35:08.171790 18451 layer_factory.hpp:172] Creating layer 'conv2_1' of type 'Convolution'
  551. I0818 13:35:08.171795 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  552. I0818 13:35:08.171828 18451 net.cpp:202] Created Layer conv2_1 (8)
  553. I0818 13:35:08.171834 18451 net.cpp:574] conv2_1 <- pool1
  554. I0818 13:35:08.171842 18451 net.cpp:544] conv2_1 -> conv2_1
  555. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  556. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  557. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  558. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  559. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  560. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  561. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  562. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  563. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  564. in math_function.cpp, line 730, tmp_value = 1048576
  565. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  566. definition of type is half.
  567. gspt r[i] = 0.0078125 (Q_frac=27
  568. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  569. In blob.cpp, blob.Q_frac= 27
  570. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  571. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  572. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  573. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  574. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  575. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  576. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  577. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  578. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  579. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  580. in math_function.cpp, line 1227, tmp_value = 0
  581. definition of type is half [FILL ZERO].
  582. gspt r[i] = 0 (Q_frac=22
  583. In blob.cpp, blob.Q_frac= 22
  584. I0818 13:35:08.270072 18451 net.cpp:262] Setting up conv2_1
  585. I0818 13:35:08.270128 18451 net.cpp:269] TRAIN Top shape for layer 8 'conv2_1' 13 256 14 14 (652288)
  586. I0818 13:35:08.270174 18451 layer_factory.hpp:172] Creating layer 'conv2_1_bn' of type 'BatchNorm'
  587. I0818 13:35:08.270182 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  588. I0818 13:35:08.270229 18451 net.cpp:202] Created Layer conv2_1_bn (9)
  589. I0818 13:35:08.270237 18451 net.cpp:574] conv2_1_bn <- conv2_1
  590. I0818 13:35:08.270256 18451 net.cpp:529] conv2_1_bn -> conv2_1 (in-place)
  591. I0818 13:35:08.270848 18451 net.cpp:262] Setting up conv2_1_bn
  592. I0818 13:35:08.270859 18451 net.cpp:269] TRAIN Top shape for layer 9 'conv2_1_bn' 13 256 14 14 (652288)
  593. I0818 13:35:08.270881 18451 layer_factory.hpp:172] Creating layer 'relu2_1' of type 'ReLU'
  594. I0818 13:35:08.270886 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  595. I0818 13:35:08.270896 18451 net.cpp:202] Created Layer relu2_1 (10)
  596. I0818 13:35:08.270902 18451 net.cpp:574] relu2_1 <- conv2_1
  597. I0818 13:35:08.270907 18451 net.cpp:529] relu2_1 -> conv2_1 (in-place)
  598. I0818 13:35:08.270917 18451 net.cpp:262] Setting up relu2_1
  599. I0818 13:35:08.270923 18451 net.cpp:269] TRAIN Top shape for layer 10 'relu2_1' 13 256 14 14 (652288)
  600. I0818 13:35:08.270929 18451 layer_factory.hpp:172] Creating layer 'conv2_2' of type 'Convolution'
  601. I0818 13:35:08.270934 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  602. I0818 13:35:08.270982 18451 net.cpp:202] Created Layer conv2_2 (11)
  603. I0818 13:35:08.271008 18451 net.cpp:574] conv2_2 <- conv2_1
  604. I0818 13:35:08.271013 18451 net.cpp:544] conv2_2 -> conv2_2
  605. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  606. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  607. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  608. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  609. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  610. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  611. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  612. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  613. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  614. in math_function.cpp, line 730, tmp_value = 1048576
  615. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  616. definition of type is half.
  617. gspt r[i] = 0.0078125 (Q_frac=27
  618. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  619. In blob.cpp, blob.Q_frac= 27
  620. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  621. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  622. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  623. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  624. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  625. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  626. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  627. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  628. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  629. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  630. in math_function.cpp, line 1227, tmp_value = 0
  631. definition of type is half [FILL ZERO].
  632. gspt r[i] = 0 (Q_frac=22
  633. In blob.cpp, blob.Q_frac= 22
  634. I0818 13:35:08.468375 18451 net.cpp:262] Setting up conv2_2
  635. I0818 13:35:08.468415 18451 net.cpp:269] TRAIN Top shape for layer 11 'conv2_2' 13 256 14 14 (652288)
  636. I0818 13:35:08.468439 18451 layer_factory.hpp:172] Creating layer 'conv2_2_bn' of type 'BatchNorm'
  637. I0818 13:35:08.468454 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  638. I0818 13:35:08.468477 18451 net.cpp:202] Created Layer conv2_2_bn (12)
  639. I0818 13:35:08.468488 18451 net.cpp:574] conv2_2_bn <- conv2_2
  640. I0818 13:35:08.468497 18451 net.cpp:529] conv2_2_bn -> conv2_2 (in-place)
  641. I0818 13:35:08.469019 18451 net.cpp:262] Setting up conv2_2_bn
  642. I0818 13:35:08.469030 18451 net.cpp:269] TRAIN Top shape for layer 12 'conv2_2_bn' 13 256 14 14 (652288)
  643. I0818 13:35:08.469043 18451 layer_factory.hpp:172] Creating layer 'relu2_2' of type 'ReLU'
  644. I0818 13:35:08.469048 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  645. I0818 13:35:08.469056 18451 net.cpp:202] Created Layer relu2_2 (13)
  646. I0818 13:35:08.469060 18451 net.cpp:574] relu2_2 <- conv2_2
  647. I0818 13:35:08.469065 18451 net.cpp:529] relu2_2 -> conv2_2 (in-place)
  648. I0818 13:35:08.469075 18451 net.cpp:262] Setting up relu2_2
  649. I0818 13:35:08.469082 18451 net.cpp:269] TRAIN Top shape for layer 13 'relu2_2' 13 256 14 14 (652288)
  650. I0818 13:35:08.469087 18451 layer_factory.hpp:172] Creating layer 'pool2' of type 'Pooling'
  651. I0818 13:35:08.469092 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  652. I0818 13:35:08.469105 18451 net.cpp:202] Created Layer pool2 (14)
  653. I0818 13:35:08.469112 18451 net.cpp:574] pool2 <- conv2_2
  654. I0818 13:35:08.469117 18451 net.cpp:544] pool2 -> pool2
  655. I0818 13:35:08.469197 18451 net.cpp:262] Setting up pool2
  656. I0818 13:35:08.469205 18451 net.cpp:269] TRAIN Top shape for layer 14 'pool2' 13 256 7 7 (163072)
  657. I0818 13:35:08.469213 18451 layer_factory.hpp:172] Creating layer 'conv3_1' of type 'Convolution'
  658. I0818 13:35:08.469218 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  659. I0818 13:35:08.469246 18451 net.cpp:202] Created Layer conv3_1 (15)
  660. I0818 13:35:08.469252 18451 net.cpp:574] conv3_1 <- pool2
  661. I0818 13:35:08.469259 18451 net.cpp:544] conv3_1 -> conv3_1
  662. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  663. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  664. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  665. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  666. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  667. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  668. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  669. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  670. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  671. in math_function.cpp, line 730, tmp_value = 1048576
  672. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  673. definition of type is half.
  674. gspt r[i] = 0.0078125 (Q_frac=27
  675. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  676. In blob.cpp, blob.Q_frac= 27
  677. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  678. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  679. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  680. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  681. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  682. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  683. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  684. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  685. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  686. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  687. in math_function.cpp, line 1227, tmp_value = 0
  688. definition of type is half [FILL ZERO].
  689. gspt r[i] = 0 (Q_frac=22
  690. In blob.cpp, blob.Q_frac= 22
  691. I0818 13:35:08.782562 18451 net.cpp:262] Setting up conv3_1
  692. I0818 13:35:08.782595 18451 net.cpp:269] TRAIN Top shape for layer 15 'conv3_1' 13 512 7 7 (326144)
  693. I0818 13:35:08.782613 18451 layer_factory.hpp:172] Creating layer 'conv3_1_bn' of type 'BatchNorm'
  694. I0818 13:35:08.782621 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  695. I0818 13:35:08.782677 18451 net.cpp:202] Created Layer conv3_1_bn (16)
  696. I0818 13:35:08.782686 18451 net.cpp:574] conv3_1_bn <- conv3_1
  697. I0818 13:35:08.782694 18451 net.cpp:529] conv3_1_bn -> conv3_1 (in-place)
  698. I0818 13:35:08.783143 18451 net.cpp:262] Setting up conv3_1_bn
  699. I0818 13:35:08.783152 18451 net.cpp:269] TRAIN Top shape for layer 16 'conv3_1_bn' 13 512 7 7 (326144)
  700. I0818 13:35:08.783176 18451 layer_factory.hpp:172] Creating layer 'relu3_1' of type 'ReLU'
  701. I0818 13:35:08.783181 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  702. I0818 13:35:08.783190 18451 net.cpp:202] Created Layer relu3_1 (17)
  703. I0818 13:35:08.783192 18451 net.cpp:574] relu3_1 <- conv3_1
  704. I0818 13:35:08.783197 18451 net.cpp:529] relu3_1 -> conv3_1 (in-place)
  705. I0818 13:35:08.783216 18451 net.cpp:262] Setting up relu3_1
  706. I0818 13:35:08.783222 18451 net.cpp:269] TRAIN Top shape for layer 17 'relu3_1' 13 512 7 7 (326144)
  707. I0818 13:35:08.783226 18451 layer_factory.hpp:172] Creating layer 'conv3_2' of type 'Convolution'
  708. I0818 13:35:08.783231 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  709. I0818 13:35:08.783254 18451 net.cpp:202] Created Layer conv3_2 (18)
  710. I0818 13:35:08.783282 18451 net.cpp:574] conv3_2 <- conv3_1
  711. I0818 13:35:08.783287 18451 net.cpp:544] conv3_2 -> conv3_2
  712. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  713. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  714. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  715. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  716. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  717. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  718. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  719. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  720. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  721. in math_function.cpp, line 730, tmp_value = 1048576
  722. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  723. definition of type is half.
  724. gspt r[i] = 0.00390625 (Q_frac=28
  725. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  726. In blob.cpp, blob.Q_frac= 28
  727. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  728. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  729. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  730. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  731. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  732. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  733. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  734. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  735. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  736. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  737. in math_function.cpp, line 1227, tmp_value = 0
  738. definition of type is half [FILL ZERO].
  739. gspt r[i] = 0 (Q_frac=22
  740. In blob.cpp, blob.Q_frac= 22
  741. I0818 13:35:09.360704 18451 net.cpp:262] Setting up conv3_2
  742. I0818 13:35:09.360775 18451 net.cpp:269] TRAIN Top shape for layer 18 'conv3_2' 13 512 7 7 (326144)
  743. I0818 13:35:09.360810 18451 layer_factory.hpp:172] Creating layer 'conv3_2_bn' of type 'BatchNorm'
  744. I0818 13:35:09.360816 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  745. I0818 13:35:09.360849 18451 net.cpp:202] Created Layer conv3_2_bn (19)
  746. I0818 13:35:09.360857 18451 net.cpp:574] conv3_2_bn <- conv3_2
  747. I0818 13:35:09.360864 18451 net.cpp:529] conv3_2_bn -> conv3_2 (in-place)
  748. I0818 13:35:09.361340 18451 net.cpp:262] Setting up conv3_2_bn
  749. I0818 13:35:09.361348 18451 net.cpp:269] TRAIN Top shape for layer 19 'conv3_2_bn' 13 512 7 7 (326144)
  750. I0818 13:35:09.361373 18451 layer_factory.hpp:172] Creating layer 'relu3_2' of type 'ReLU'
  751. I0818 13:35:09.361378 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  752. I0818 13:35:09.361385 18451 net.cpp:202] Created Layer relu3_2 (20)
  753. I0818 13:35:09.361389 18451 net.cpp:574] relu3_2 <- conv3_2
  754. I0818 13:35:09.361393 18451 net.cpp:529] relu3_2 -> conv3_2 (in-place)
  755. I0818 13:35:09.361418 18451 net.cpp:262] Setting up relu3_2
  756. I0818 13:35:09.361423 18451 net.cpp:269] TRAIN Top shape for layer 20 'relu3_2' 13 512 7 7 (326144)
  757. I0818 13:35:09.361428 18451 layer_factory.hpp:172] Creating layer 'pool3' of type 'Pooling'
  758. I0818 13:35:09.361431 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  759. I0818 13:35:09.361459 18451 net.cpp:202] Created Layer pool3 (21)
  760. I0818 13:35:09.361464 18451 net.cpp:574] pool3 <- conv3_2
  761. I0818 13:35:09.361467 18451 net.cpp:544] pool3 -> pool3
  762. I0818 13:35:09.361546 18451 net.cpp:262] Setting up pool3
  763. I0818 13:35:09.361557 18451 net.cpp:269] TRAIN Top shape for layer 21 'pool3' 13 512 4 4 (106496)
  764. I0818 13:35:09.361562 18451 layer_factory.hpp:172] Creating layer 'ip1' of type 'InnerProduct'
  765. I0818 13:35:09.361572 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  766. I0818 13:35:09.361610 18451 net.cpp:202] Created Layer ip1 (22)
  767. I0818 13:35:09.361618 18451 net.cpp:574] ip1 <- pool3
  768. I0818 13:35:09.361622 18451 net.cpp:544] ip1 -> ip1
  769. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  770. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  771. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  772. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  773. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  774. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  775. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  776. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  777. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  778. in math_function.cpp, line 730, tmp_value = 1048576
  779. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  780. definition of type is half.
  781. gspt r[i] = 0.00390625 (Q_frac=28
  782. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  783. In blob.cpp, blob.Q_frac= 28
  784. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  785. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  786. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  787. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  788. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  789. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  790. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  791. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  792. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  793. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  794. in math_function.cpp, line 1227, tmp_value = 0
  795. definition of type is half [FILL ZERO].
  796. gspt r[i] = 0 (Q_frac=22
  797. In blob.cpp, blob.Q_frac= 22
  798. I0818 13:35:11.366432 18451 net.cpp:262] Setting up ip1
  799. I0818 13:35:11.366459 18451 net.cpp:269] TRAIN Top shape for layer 22 'ip1' 13 1024 (13312)
  800. I0818 13:35:11.366489 18451 layer_factory.hpp:172] Creating layer 'relu4' of type 'ReLU'
  801. I0818 13:35:11.366497 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  802. I0818 13:35:11.366523 18451 net.cpp:202] Created Layer relu4 (23)
  803. I0818 13:35:11.366528 18451 net.cpp:574] relu4 <- ip1
  804. I0818 13:35:11.366536 18451 net.cpp:529] relu4 -> ip1 (in-place)
  805. I0818 13:35:11.366550 18451 net.cpp:262] Setting up relu4
  806. I0818 13:35:11.366554 18451 net.cpp:269] TRAIN Top shape for layer 23 'relu4' 13 1024 (13312)
  807. I0818 13:35:11.366559 18451 layer_factory.hpp:172] Creating layer 'drop7' of type 'Dropout'
  808. I0818 13:35:11.366562 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  809. I0818 13:35:11.366585 18451 net.cpp:202] Created Layer drop7 (24)
  810. I0818 13:35:11.366588 18451 net.cpp:574] drop7 <- ip1
  811. I0818 13:35:11.366591 18451 net.cpp:529] drop7 -> ip1 (in-place)
  812. I0818 13:35:11.377153 18451 net.cpp:262] Setting up drop7
  813. I0818 13:35:11.377184 18451 net.cpp:269] TRAIN Top shape for layer 24 'drop7' 13 1024 (13312)
  814. I0818 13:35:11.377190 18451 layer_factory.hpp:172] Creating layer 'ip2' of type 'InnerProduct'
  815. I0818 13:35:11.377195 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  816. I0818 13:35:11.377223 18451 net.cpp:202] Created Layer ip2 (25)
  817. I0818 13:35:11.377241 18451 net.cpp:574] ip2 <- ip1
  818. I0818 13:35:11.377247 18451 net.cpp:544] ip2 -> ip2
  819. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  820. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  821. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  822. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  823. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  824. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  825. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  826. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  827. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  828. in math_function.cpp, line 730, tmp_value = 1048576
  829. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  830. definition of type is half.
  831. gspt r[i] = 0.0078125 (Q_frac=27
  832. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  833. In blob.cpp, blob.Q_frac= 27
  834. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  835. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  836. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  837. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  838. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  839. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  840. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  841. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  842. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  843. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  844. in math_function.cpp, line 1227, tmp_value = 0
  845. definition of type is half [FILL ZERO].
  846. gspt r[i] = 0 (Q_frac=22
  847. In blob.cpp, blob.Q_frac= 22
  848. I0818 13:35:11.386552 18451 net.cpp:262] Setting up ip2
  849. I0818 13:35:11.386592 18451 net.cpp:269] TRAIN Top shape for layer 25 'ip2' 13 10 (130)
  850. I0818 13:35:11.386652 18451 layer_factory.hpp:172] Creating layer 'loss' of type 'SoftmaxWithLoss'
  851. I0818 13:35:11.386662 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  852. I0818 13:35:11.386680 18451 net.cpp:202] Created Layer loss (26)
  853. I0818 13:35:11.386687 18451 net.cpp:574] loss <- ip2
  854. I0818 13:35:11.386693 18451 net.cpp:574] loss <- label
  855. I0818 13:35:11.386704 18451 net.cpp:544] loss -> loss
  856. I0818 13:35:11.386905 18451 net.cpp:262] Setting up loss
  857. I0818 13:35:11.386915 18451 net.cpp:269] TRAIN Top shape for layer 26 'loss' (1)
  858. I0818 13:35:11.386919 18451 net.cpp:273] with loss weight 1
  859. I0818 13:35:11.386926 18451 net.cpp:338] loss needs backward computation.
  860. I0818 13:35:11.386931 18451 net.cpp:338] ip2 needs backward computation.
  861. I0818 13:35:11.386936 18451 net.cpp:338] drop7 needs backward computation.
  862. I0818 13:35:11.386940 18451 net.cpp:338] relu4 needs backward computation.
  863. I0818 13:35:11.386943 18451 net.cpp:338] ip1 needs backward computation.
  864. I0818 13:35:11.386947 18451 net.cpp:338] pool3 needs backward computation.
  865. I0818 13:35:11.386952 18451 net.cpp:338] relu3_2 needs backward computation.
  866. I0818 13:35:11.386956 18451 net.cpp:338] conv3_2_bn needs backward computation.
  867. I0818 13:35:11.386960 18451 net.cpp:338] conv3_2 needs backward computation.
  868. I0818 13:35:11.386965 18451 net.cpp:338] relu3_1 needs backward computation.
  869. I0818 13:35:11.386968 18451 net.cpp:338] conv3_1_bn needs backward computation.
  870. I0818 13:35:11.386972 18451 net.cpp:338] conv3_1 needs backward computation.
  871. I0818 13:35:11.386976 18451 net.cpp:338] pool2 needs backward computation.
  872. I0818 13:35:11.386981 18451 net.cpp:338] relu2_2 needs backward computation.
  873. I0818 13:35:11.386984 18451 net.cpp:338] conv2_2_bn needs backward computation.
  874. I0818 13:35:11.386988 18451 net.cpp:338] conv2_2 needs backward computation.
  875. I0818 13:35:11.386992 18451 net.cpp:338] relu2_1 needs backward computation.
  876. I0818 13:35:11.386996 18451 net.cpp:338] conv2_1_bn needs backward computation.
  877. I0818 13:35:11.387001 18451 net.cpp:338] conv2_1 needs backward computation.
  878. I0818 13:35:11.387004 18451 net.cpp:338] pool1 needs backward computation.
  879. I0818 13:35:11.387008 18451 net.cpp:338] relu1_2 needs backward computation.
  880. I0818 13:35:11.387012 18451 net.cpp:338] conv1_2_bn needs backward computation.
  881. I0818 13:35:11.387017 18451 net.cpp:338] conv1_2 needs backward computation.
  882. I0818 13:35:11.387035 18451 net.cpp:338] relu1_1 needs backward computation.
  883. I0818 13:35:11.387039 18451 net.cpp:338] conv1_1_bn needs backward computation.
  884. I0818 13:35:11.387042 18451 net.cpp:338] conv1_1 needs backward computation.
  885. I0818 13:35:11.387050 18451 net.cpp:340] cifar does not need backward computation.
  886. I0818 13:35:11.387054 18451 net.cpp:382] This network produces output loss
  887. I0818 13:35:11.387097 18451 net.cpp:405] Top memory (TRAIN) required for data: 28728836 diff: 28790016
  888. I0818 13:35:11.387104 18451 net.cpp:408] Bottom memory (TRAIN) required for data: 28728832 diff: 28790012
  889. I0818 13:35:11.387109 18451 net.cpp:411] Shared (in-place) memory (TRAIN) by data: 18317312 diff: 18317312
  890. I0818 13:35:11.387111 18451 net.cpp:414] Parameters memory (TRAIN) required for data: 25966916 diff: 25966916
  891. I0818 13:35:11.387115 18451 net.cpp:417] Parameters shared memory (TRAIN) by data: 0 diff: 0
  892. I0818 13:35:11.387118 18451 net.cpp:423] Network initialization done.
  893. I0818 13:35:11.387629 18451 solver.cpp:174] Creating test net (#0) specified by net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  894. I0818 13:35:11.387697 18451 net.cpp:459] The NetState phase (1) differed from the phase (0) specified by a rule in layer cifar
  895. W0818 13:35:11.387759 18451 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  896. I0818 13:35:11.387956 18451 net.cpp:83] Initializing net from parameters:
  897. name: "CIFAR10_full_vgg7_BN"
  898. state {
  899. phase: TEST
  900. }
  901. default_forward_type: FLOAT16
  902. default_backward_type: FLOAT16
  903. default_forward_math: FLOAT16
  904. default_backward_math: FLOAT16
  905. global_grad_scale: 2000
  906. layer {
  907. name: "cifar"
  908. type: "Data"
  909. top: "data"
  910. top: "label"
  911. include {
  912. phase: TEST
  913. }
  914. transform_param {
  915. scale: 0.00390625
  916. mean_file: "examples/cifar10/mean.binaryproto"
  917. }
  918. data_param {
  919. source: "examples/cifar10/cifar10_test_lmdb"
  920. batch_size: 13
  921. backend: LMDB
  922. }
  923. }
  924. layer {
  925. name: "conv1_1"
  926. type: "Convolution"
  927. bottom: "data"
  928. top: "conv1_1"
  929. param {
  930. lr_mult: 1
  931. }
  932. param {
  933. lr_mult: 2
  934. }
  935. convolution_param {
  936. num_output: 128
  937. pad: 1
  938. kernel_size: 3
  939. stride: 1
  940. weight_filler {
  941. type: "gspt_xavier"
  942. }
  943. bias_filler {
  944. type: "gspt_constant"
  945. }
  946. }
  947. }
  948. layer {
  949. name: "conv1_1_bn"
  950. type: "BatchNorm"
  951. bottom: "conv1_1"
  952. top: "conv1_1"
  953. batch_norm_param {
  954. use_global_stats: false
  955. }
  956. }
  957. layer {
  958. name: "relu1_1"
  959. type: "ReLU"
  960. bottom: "conv1_1"
  961. top: "conv1_1"
  962. }
  963. layer {
  964. name: "conv1_2"
  965. type: "Convolution"
  966. bottom: "conv1_1"
  967. top: "conv1_2"
  968. param {
  969. lr_mult: 1
  970. }
  971. param {
  972. lr_mult: 2
  973. }
  974. convolution_param {
  975. num_output: 128
  976. pad: 1
  977. kernel_size: 3
  978. stride: 1
  979. weight_filler {
  980. type: "gspt_xavier"
  981. }
  982. bias_filler {
  983. type: "gspt_constant"
  984. }
  985. }
  986. }
  987. layer {
  988. name: "conv1_2_bn"
  989. type: "BatchNorm"
  990. bottom: "conv1_2"
  991. top: "conv1_2"
  992. batch_norm_param {
  993. use_global_stats: false
  994. }
  995. }
  996. layer {
  997. name: "relu1_2"
  998. type: "ReLU"
  999. bottom: "conv1_2"
  1000. top: "conv1_2"
  1001. }
  1002. layer {
  1003. name: "pool1"
  1004. type: "Pooling"
  1005. bottom: "conv1_2"
  1006. top: "pool1"
  1007. pooling_param {
  1008. pool: MAX
  1009. kernel_size: 2
  1010. stride: 2
  1011. }
  1012. }
  1013. layer {
  1014. name: "conv2_1"
  1015. type: "Convolution"
  1016. bottom: "pool1"
  1017. top: "conv2_1"
  1018. param {
  1019. lr_mult: 1
  1020. }
  1021. param {
  1022. lr_mult: 2
  1023. }
  1024. convolution_param {
  1025. num_output: 256
  1026. pad: 1
  1027. kernel_size: 3
  1028. stride: 1
  1029. weight_filler {
  1030. type: "gspt_xavier"
  1031. }
  1032. bias_filler {
  1033. type: "gspt_constant"
  1034. }
  1035. }
  1036. }
  1037. layer {
  1038. name: "conv2_1_bn"
  1039. type: "BatchNorm"
  1040. bottom: "conv2_1"
  1041. top: "conv2_1"
  1042. batch_norm_param {
  1043. use_global_stats: false
  1044. }
  1045. }
  1046. layer {
  1047. name: "relu2_1"
  1048. type: "ReLU"
  1049. bottom: "conv2_1"
  1050. top: "conv2_1"
  1051. }
  1052. layer {
  1053. name: "conv2_2"
  1054. type: "Convolution"
  1055. bottom: "conv2_1"
  1056. top: "conv2_2"
  1057. param {
  1058. lr_mult: 1
  1059. }
  1060. param {
  1061. lr_mult: 2
  1062. }
  1063. convolution_param {
  1064. num_output: 256
  1065. pad: 1
  1066. kernel_size: 3
  1067. stride: 1
  1068. weight_filler {
  1069. type: "gspt_xavier"
  1070. }
  1071. bias_filler {
  1072. type: "gspt_constant"
  1073. }
  1074. }
  1075. }
  1076. layer {
  1077. name: "conv2_2_bn"
  1078. type: "BatchNorm"
  1079. bottom: "conv2_2"
  1080. top: "conv2_2"
  1081. batch_norm_param {
  1082. use_global_stats: false
  1083. }
  1084. }
  1085. layer {
  1086. name: "relu2_2"
  1087. type: "ReLU"
  1088. bottom: "conv2_2"
  1089. top: "conv2_2"
  1090. }
  1091. layer {
  1092. name: "pool2"
  1093. type: "Pooling"
  1094. bottom: "conv2_2"
  1095. top: "pool2"
  1096. pooling_param {
  1097. pool: MAX
  1098. kernel_size: 2
  1099. stride: 2
  1100. }
  1101. }
  1102. layer {
  1103. name: "conv3_1"
  1104. type: "Convolution"
  1105. bottom: "pool2"
  1106. top: "conv3_1"
  1107. param {
  1108. lr_mult: 1
  1109. }
  1110. param {
  1111. lr_mult: 2
  1112. }
  1113. convolution_param {
  1114. num_output: 512
  1115. pad: 1
  1116. kernel_size: 3
  1117. stride: 1
  1118. weight_filler {
  1119. type: "gspt_xavier"
  1120. }
  1121. bias_filler {
  1122. type: "gspt_constant"
  1123. }
  1124. }
  1125. }
  1126. layer {
  1127. name: "conv3_1_bn"
  1128. type: "BatchNorm"
  1129. bottom: "conv3_1"
  1130. top: "conv3_1"
  1131. batch_norm_param {
  1132. use_global_stats: false
  1133. }
  1134. }
  1135. layer {
  1136. name: "relu3_1"
  1137. type: "ReLU"
  1138. bottom: "conv3_1"
  1139. top: "conv3_1"
  1140. }
  1141. layer {
  1142. name: "conv3_2"
  1143. type: "Convolution"
  1144. bottom: "conv3_1"
  1145. top: "conv3_2"
  1146. param {
  1147. lr_mult: 1
  1148. }
  1149. param {
  1150. lr_mult: 2
  1151. }
  1152. convolution_param {
  1153. num_output: 512
  1154. pad: 1
  1155. kernel_size: 3
  1156. stride: 1
  1157. weight_filler {
  1158. type: "gspt_xavier"
  1159. }
  1160. bias_filler {
  1161. type: "gspt_constant"
  1162. }
  1163. }
  1164. }
  1165. layer {
  1166. name: "conv3_2_bn"
  1167. type: "BatchNorm"
  1168. bottom: "conv3_2"
  1169. top: "conv3_2"
  1170. batch_norm_param {
  1171. use_global_stats: false
  1172. }
  1173. }
  1174. layer {
  1175. name: "relu3_2"
  1176. type: "ReLU"
  1177. bottom: "conv3_2"
  1178. top: "conv3_2"
  1179. }
  1180. layer {
  1181. name: "pool3"
  1182. type: "Pooling"
  1183. bottom: "conv3_2"
  1184. top: "pool3"
  1185. pooling_param {
  1186. pool: MAX
  1187. kernel_size: 2
  1188. stride: 2
  1189. }
  1190. }
  1191. layer {
  1192. name: "ip1"
  1193. type: "InnerProduct"
  1194. bottom: "pool3"
  1195. top: "ip1"
  1196. param {
  1197. lr_mult: 1
  1198. decay_mult: 0
  1199. }
  1200. param {
  1201. lr_mult: 2
  1202. decay_mult: 0
  1203. }
  1204. inner_product_param {
  1205. num_output: 1024
  1206. weight_filler {
  1207. type: "gspt_xavier"
  1208. }
  1209. bias_filler {
  1210. type: "gspt_constant"
  1211. }
  1212. }
  1213. }
  1214. layer {
  1215. name: "relu4"
  1216. type: "ReLU"
  1217. bottom: "ip1"
  1218. top: "ip1"
  1219. }
  1220. layer {
  1221. name: "drop7"
  1222. type: "Dropout"
  1223. bottom: "ip1"
  1224. top: "ip1"
  1225. dropout_param {
  1226. dropout_ratio: 0.5
  1227. }
  1228. }
  1229. layer {
  1230. name: "ip2"
  1231. type: "InnerProduct"
  1232. bottom: "ip1"
  1233. top: "ip2"
  1234. param {
  1235. lr_mult: 1
  1236. decay_mult: 0
  1237. }
  1238. param {
  1239. lr_mult: 2
  1240. decay_mult: 0
  1241. }
  1242. inner_product_param {
  1243. num_output: 10
  1244. weight_filler {
  1245. type: "gspt_xavier"
  1246. }
  1247. bias_filler {
  1248. type: "gspt_constant"
  1249. }
  1250. }
  1251. }
  1252. layer {
  1253. name: "accuracy"
  1254. type: "Accuracy"
  1255. bottom: "ip2"
  1256. bottom: "label"
  1257. top: "accuracy"
  1258. include {
  1259. phase: TEST
  1260. }
  1261. }
  1262. layer {
  1263. name: "loss"
  1264. type: "SoftmaxWithLoss"
  1265. bottom: "ip2"
  1266. bottom: "label"
  1267. top: "loss"
  1268. }
  1269. I0818 13:35:11.388160 18451 layer_factory.hpp:172] Creating layer 'cifar' of type 'Data'
  1270. I0818 13:35:11.388168 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT Fmath:FLOAT16 Bmath:FLOAT16
  1271. I0818 13:35:11.388202 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 0
  1272. I0818 13:35:11.388378 18451 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  1273. I0818 13:35:11.390437 18451 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  1274. I0818 13:35:11.390645 18451 net.cpp:202] Created Layer cifar (0)
  1275. I0818 13:35:11.390682 18451 net.cpp:544] cifar -> data
  1276. I0818 13:35:11.390698 18451 net.cpp:544] cifar -> label
  1277. I0818 13:35:11.390710 18451 data_reader.cpp:59] Data Reader threads: 1, out queues: 1, depth: 13
  1278. I0818 13:35:11.390727 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 0
  1279. I0818 13:35:11.392611 18511 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_test_lmdb
  1280. I0818 13:35:11.392872 18451 data_layer.cpp:200] (n0.d0.r0) Output data size: 13, 3, 32, 32
  1281. I0818 13:35:11.392892 18451 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  1282. I0818 13:35:11.393024 18451 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  1283. I0818 13:35:11.393144 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 0
  1284. I0818 13:35:11.393193 18451 net.cpp:262] Setting up cifar
  1285. I0818 13:35:11.393204 18451 net.cpp:269] TEST Top shape for layer 0 'cifar' 13 3 32 32 (39936)
  1286. I0818 13:35:11.393209 18451 net.cpp:269] TEST Top shape for layer 0 'cifar' 13 (13)
  1287. I0818 13:35:11.393214 18451 layer_factory.hpp:172] Creating layer 'label_cifar_1_split' of type 'Split'
  1288. I0818 13:35:11.393219 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1289. I0818 13:35:11.393250 18451 net.cpp:202] Created Layer label_cifar_1_split (1)
  1290. I0818 13:35:11.393258 18451 net.cpp:574] label_cifar_1_split <- label
  1291. I0818 13:35:11.393265 18451 net.cpp:544] label_cifar_1_split -> label_cifar_1_split_0
  1292. I0818 13:35:11.393275 18451 net.cpp:544] label_cifar_1_split -> label_cifar_1_split_1
  1293. I0818 13:35:11.395053 18512 data_layer.cpp:105] (n0.d0.r0) Parser threads: 1
  1294. I0818 13:35:11.395063 18451 net.cpp:262] Setting up label_cifar_1_split
  1295. I0818 13:35:11.395092 18512 data_layer.cpp:107] (n0.d0.r0) Transformer threads: 1
  1296. I0818 13:35:11.395112 18451 net.cpp:269] TEST Top shape for layer 1 'label_cifar_1_split' 13 (13)
  1297. I0818 13:35:11.395119 18451 net.cpp:269] TEST Top shape for layer 1 'label_cifar_1_split' 13 (13)
  1298. I0818 13:35:11.395124 18451 layer_factory.hpp:172] Creating layer 'conv1_1' of type 'Convolution'
  1299. I0818 13:35:11.395128 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1300. I0818 13:35:11.395171 18451 net.cpp:202] Created Layer conv1_1 (2)
  1301. I0818 13:35:11.395180 18451 net.cpp:574] conv1_1 <- data
  1302. I0818 13:35:11.395200 18451 net.cpp:544] conv1_1 -> conv1_1
  1303. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1304. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1305. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1306. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1307. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1308. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1309. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1310. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1311. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1312. in math_function.cpp, line 730, tmp_value = 1048576
  1313. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1314. definition of type is half.
  1315. gspt r[i] = 0.0625 (Q_frac=24
  1316. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1317. In blob.cpp, blob.Q_frac= 24
  1318. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1319. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1320. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1321. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1322. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1323. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1324. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1325. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1326. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1327. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1328. in math_function.cpp, line 1227, tmp_value = 0
  1329. definition of type is half [FILL ZERO].
  1330. gspt r[i] = 0 (Q_frac=22
  1331. In blob.cpp, blob.Q_frac= 22
  1332. I0818 13:35:11.400393 18451 net.cpp:262] Setting up conv1_1
  1333. I0818 13:35:11.400411 18451 net.cpp:269] TEST Top shape for layer 2 'conv1_1' 13 128 32 32 (1703936)
  1334. I0818 13:35:11.400426 18451 layer_factory.hpp:172] Creating layer 'conv1_1_bn' of type 'BatchNorm'
  1335. I0818 13:35:11.400434 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1336. I0818 13:35:11.400451 18451 net.cpp:202] Created Layer conv1_1_bn (3)
  1337. I0818 13:35:11.400458 18451 net.cpp:574] conv1_1_bn <- conv1_1
  1338. I0818 13:35:11.400465 18451 net.cpp:529] conv1_1_bn -> conv1_1 (in-place)
  1339. I0818 13:35:11.401283 18451 net.cpp:262] Setting up conv1_1_bn
  1340. I0818 13:35:11.401310 18451 net.cpp:269] TEST Top shape for layer 3 'conv1_1_bn' 13 128 32 32 (1703936)
  1341. I0818 13:35:11.401334 18451 layer_factory.hpp:172] Creating layer 'relu1_1' of type 'ReLU'
  1342. I0818 13:35:11.401338 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1343. I0818 13:35:11.401365 18451 net.cpp:202] Created Layer relu1_1 (4)
  1344. I0818 13:35:11.401371 18451 net.cpp:574] relu1_1 <- conv1_1
  1345. I0818 13:35:11.401377 18451 net.cpp:529] relu1_1 -> conv1_1 (in-place)
  1346. I0818 13:35:11.401404 18451 net.cpp:262] Setting up relu1_1
  1347. I0818 13:35:11.401412 18451 net.cpp:269] TEST Top shape for layer 4 'relu1_1' 13 128 32 32 (1703936)
  1348. I0818 13:35:11.401417 18451 layer_factory.hpp:172] Creating layer 'conv1_2' of type 'Convolution'
  1349. I0818 13:35:11.401422 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1350. I0818 13:35:11.401490 18451 net.cpp:202] Created Layer conv1_2 (5)
  1351. I0818 13:35:11.401496 18451 net.cpp:574] conv1_2 <- conv1_1
  1352. I0818 13:35:11.401501 18451 net.cpp:544] conv1_2 -> conv1_2
  1353. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1354. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1355. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1356. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1357. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1358. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1359. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1360. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1361. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1362. in math_function.cpp, line 730, tmp_value = 1048576
  1363. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1364. definition of type is half.
  1365. gspt r[i] = 0.0078125 (Q_frac=27
  1366. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1367. In blob.cpp, blob.Q_frac= 27
  1368. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1369. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1370. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1371. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1372. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1373. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1374. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1375. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1376. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1377. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1378. in math_function.cpp, line 1227, tmp_value = 0
  1379. definition of type is half [FILL ZERO].
  1380. gspt r[i] = 0 (Q_frac=22
  1381. In blob.cpp, blob.Q_frac= 22
  1382. I0818 13:35:11.441438 18451 net.cpp:262] Setting up conv1_2
  1383. I0818 13:35:11.441468 18451 net.cpp:269] TEST Top shape for layer 5 'conv1_2' 13 128 32 32 (1703936)
  1384. I0818 13:35:11.441478 18451 layer_factory.hpp:172] Creating layer 'conv1_2_bn' of type 'BatchNorm'
  1385. I0818 13:35:11.441481 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1386. I0818 13:35:11.441511 18451 net.cpp:202] Created Layer conv1_2_bn (6)
  1387. I0818 13:35:11.441516 18451 net.cpp:574] conv1_2_bn <- conv1_2
  1388. I0818 13:35:11.441522 18451 net.cpp:529] conv1_2_bn -> conv1_2 (in-place)
  1389. I0818 13:35:11.442016 18451 net.cpp:262] Setting up conv1_2_bn
  1390. I0818 13:35:11.442025 18451 net.cpp:269] TEST Top shape for layer 6 'conv1_2_bn' 13 128 32 32 (1703936)
  1391. I0818 13:35:11.442056 18451 layer_factory.hpp:172] Creating layer 'relu1_2' of type 'ReLU'
  1392. I0818 13:35:11.442061 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1393. I0818 13:35:11.442066 18451 net.cpp:202] Created Layer relu1_2 (7)
  1394. I0818 13:35:11.442070 18451 net.cpp:574] relu1_2 <- conv1_2
  1395. I0818 13:35:11.442090 18451 net.cpp:529] relu1_2 -> conv1_2 (in-place)
  1396. I0818 13:35:11.442096 18451 net.cpp:262] Setting up relu1_2
  1397. I0818 13:35:11.442101 18451 net.cpp:269] TEST Top shape for layer 7 'relu1_2' 13 128 32 32 (1703936)
  1398. I0818 13:35:11.442106 18451 layer_factory.hpp:172] Creating layer 'pool1' of type 'Pooling'
  1399. I0818 13:35:11.442126 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1400. I0818 13:35:11.442142 18451 net.cpp:202] Created Layer pool1 (8)
  1401. I0818 13:35:11.442147 18451 net.cpp:574] pool1 <- conv1_2
  1402. I0818 13:35:11.442152 18451 net.cpp:544] pool1 -> pool1
  1403. I0818 13:35:11.442221 18451 net.cpp:262] Setting up pool1
  1404. I0818 13:35:11.442230 18451 net.cpp:269] TEST Top shape for layer 8 'pool1' 13 128 16 16 (425984)
  1405. I0818 13:35:11.442234 18451 layer_factory.hpp:172] Creating layer 'conv2_1' of type 'Convolution'
  1406. I0818 13:35:11.442239 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1407. I0818 13:35:11.442262 18451 net.cpp:202] Created Layer conv2_1 (9)
  1408. I0818 13:35:11.442267 18451 net.cpp:574] conv2_1 <- pool1
  1409. I0818 13:35:11.442272 18451 net.cpp:544] conv2_1 -> conv2_1
  1410. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1411. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1412. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1413. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1414. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1415. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1416. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1417. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1418. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1419. in math_function.cpp, line 730, tmp_value = 1048576
  1420. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1421. definition of type is half.
  1422. gspt r[i] = 0.0078125 (Q_frac=27
  1423. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1424. In blob.cpp, blob.Q_frac= 27
  1425. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1426. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1427. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1428. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1429. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1430. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1431. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1432. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1433. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1434. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1435. in math_function.cpp, line 1227, tmp_value = 0
  1436. definition of type is half [FILL ZERO].
  1437. gspt r[i] = 0 (Q_frac=22
  1438. In blob.cpp, blob.Q_frac= 22
  1439. I0818 13:35:11.519840 18451 net.cpp:262] Setting up conv2_1
  1440. I0818 13:35:11.519876 18451 net.cpp:269] TEST Top shape for layer 9 'conv2_1' 13 256 16 16 (851968)
  1441. I0818 13:35:11.519898 18451 layer_factory.hpp:172] Creating layer 'conv2_1_bn' of type 'BatchNorm'
  1442. I0818 13:35:11.519908 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1443. I0818 13:35:11.519929 18451 net.cpp:202] Created Layer conv2_1_bn (10)
  1444. I0818 13:35:11.519938 18451 net.cpp:574] conv2_1_bn <- conv2_1
  1445. I0818 13:35:11.519948 18451 net.cpp:529] conv2_1_bn -> conv2_1 (in-place)
  1446. I0818 13:35:11.520458 18451 net.cpp:262] Setting up conv2_1_bn
  1447. I0818 13:35:11.520468 18451 net.cpp:269] TEST Top shape for layer 10 'conv2_1_bn' 13 256 16 16 (851968)
  1448. I0818 13:35:11.520491 18451 layer_factory.hpp:172] Creating layer 'relu2_1' of type 'ReLU'
  1449. I0818 13:35:11.520496 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1450. I0818 13:35:11.520503 18451 net.cpp:202] Created Layer relu2_1 (11)
  1451. I0818 13:35:11.520507 18451 net.cpp:574] relu2_1 <- conv2_1
  1452. I0818 13:35:11.520512 18451 net.cpp:529] relu2_1 -> conv2_1 (in-place)
  1453. I0818 13:35:11.520535 18451 net.cpp:262] Setting up relu2_1
  1454. I0818 13:35:11.520561 18451 net.cpp:269] TEST Top shape for layer 11 'relu2_1' 13 256 16 16 (851968)
  1455. I0818 13:35:11.520567 18451 layer_factory.hpp:172] Creating layer 'conv2_2' of type 'Convolution'
  1456. I0818 13:35:11.520572 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1457. I0818 13:35:11.520597 18451 net.cpp:202] Created Layer conv2_2 (12)
  1458. I0818 13:35:11.520602 18451 net.cpp:574] conv2_2 <- conv2_1
  1459. I0818 13:35:11.520607 18451 net.cpp:544] conv2_2 -> conv2_2
  1460. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1461. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1462. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1463. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1464. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1465. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1466. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1467. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1468. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1469. in math_function.cpp, line 730, tmp_value = 1048576
  1470. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1471. definition of type is half.
  1472. gspt r[i] = 0.0078125 (Q_frac=27
  1473. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1474. In blob.cpp, blob.Q_frac= 27
  1475. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1476. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1477. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1478. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1479. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1480. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1481. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1482. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1483. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1484. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1485. in math_function.cpp, line 1227, tmp_value = 0
  1486. definition of type is half [FILL ZERO].
  1487. gspt r[i] = 0 (Q_frac=22
  1488. In blob.cpp, blob.Q_frac= 22
  1489. I0818 13:35:11.664553 18451 net.cpp:262] Setting up conv2_2
  1490. I0818 13:35:11.664587 18451 net.cpp:269] TEST Top shape for layer 12 'conv2_2' 13 256 16 16 (851968)
  1491. I0818 13:35:11.664636 18451 layer_factory.hpp:172] Creating layer 'conv2_2_bn' of type 'BatchNorm'
  1492. I0818 13:35:11.664659 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1493. I0818 13:35:11.664695 18451 net.cpp:202] Created Layer conv2_2_bn (13)
  1494. I0818 13:35:11.664701 18451 net.cpp:574] conv2_2_bn <- conv2_2
  1495. I0818 13:35:11.664710 18451 net.cpp:529] conv2_2_bn -> conv2_2 (in-place)
  1496. I0818 13:35:11.665282 18451 net.cpp:262] Setting up conv2_2_bn
  1497. I0818 13:35:11.665292 18451 net.cpp:269] TEST Top shape for layer 13 'conv2_2_bn' 13 256 16 16 (851968)
  1498. I0818 13:35:11.665316 18451 layer_factory.hpp:172] Creating layer 'relu2_2' of type 'ReLU'
  1499. I0818 13:35:11.665326 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1500. I0818 13:35:11.665335 18451 net.cpp:202] Created Layer relu2_2 (14)
  1501. I0818 13:35:11.665339 18451 net.cpp:574] relu2_2 <- conv2_2
  1502. I0818 13:35:11.665343 18451 net.cpp:529] relu2_2 -> conv2_2 (in-place)
  1503. I0818 13:35:11.665351 18451 net.cpp:262] Setting up relu2_2
  1504. I0818 13:35:11.665356 18451 net.cpp:269] TEST Top shape for layer 14 'relu2_2' 13 256 16 16 (851968)
  1505. I0818 13:35:11.665374 18451 layer_factory.hpp:172] Creating layer 'pool2' of type 'Pooling'
  1506. I0818 13:35:11.665396 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1507. I0818 13:35:11.665411 18451 net.cpp:202] Created Layer pool2 (15)
  1508. I0818 13:35:11.665416 18451 net.cpp:574] pool2 <- conv2_2
  1509. I0818 13:35:11.665423 18451 net.cpp:544] pool2 -> pool2
  1510. I0818 13:35:11.665514 18451 net.cpp:262] Setting up pool2
  1511. I0818 13:35:11.665527 18451 net.cpp:269] TEST Top shape for layer 15 'pool2' 13 256 8 8 (212992)
  1512. I0818 13:35:11.665532 18451 layer_factory.hpp:172] Creating layer 'conv3_1' of type 'Convolution'
  1513. I0818 13:35:11.665537 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1514. I0818 13:35:11.665565 18451 net.cpp:202] Created Layer conv3_1 (16)
  1515. I0818 13:35:11.665570 18451 net.cpp:574] conv3_1 <- pool2
  1516. I0818 13:35:11.665575 18451 net.cpp:544] conv3_1 -> conv3_1
  1517. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1518. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1519. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1520. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1521. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1522. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1523. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1524. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1525. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1526. in math_function.cpp, line 730, tmp_value = 1048576
  1527. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1528. definition of type is half.
  1529. gspt r[i] = 0.0078125 (Q_frac=27
  1530. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1531. In blob.cpp, blob.Q_frac= 27
  1532. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1533. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1534. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1535. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1536. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1537. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1538. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1539. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1540. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1541. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1542. in math_function.cpp, line 1227, tmp_value = 0
  1543. definition of type is half [FILL ZERO].
  1544. gspt r[i] = 0 (Q_frac=22
  1545. In blob.cpp, blob.Q_frac= 22
  1546. I0818 13:35:11.971582 18451 net.cpp:262] Setting up conv3_1
  1547. I0818 13:35:11.971664 18451 net.cpp:269] TEST Top shape for layer 16 'conv3_1' 13 512 8 8 (425984)
  1548. I0818 13:35:11.971696 18451 layer_factory.hpp:172] Creating layer 'conv3_1_bn' of type 'BatchNorm'
  1549. I0818 13:35:11.971719 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1550. I0818 13:35:11.971740 18451 net.cpp:202] Created Layer conv3_1_bn (17)
  1551. I0818 13:35:11.971750 18451 net.cpp:574] conv3_1_bn <- conv3_1
  1552. I0818 13:35:11.971757 18451 net.cpp:529] conv3_1_bn -> conv3_1 (in-place)
  1553. I0818 13:35:11.972296 18451 net.cpp:262] Setting up conv3_1_bn
  1554. I0818 13:35:11.972306 18451 net.cpp:269] TEST Top shape for layer 17 'conv3_1_bn' 13 512 8 8 (425984)
  1555. I0818 13:35:11.972331 18451 layer_factory.hpp:172] Creating layer 'relu3_1' of type 'ReLU'
  1556. I0818 13:35:11.972337 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1557. I0818 13:35:11.972345 18451 net.cpp:202] Created Layer relu3_1 (18)
  1558. I0818 13:35:11.972350 18451 net.cpp:574] relu3_1 <- conv3_1
  1559. I0818 13:35:11.972354 18451 net.cpp:529] relu3_1 -> conv3_1 (in-place)
  1560. I0818 13:35:11.972378 18451 net.cpp:262] Setting up relu3_1
  1561. I0818 13:35:11.972383 18451 net.cpp:269] TEST Top shape for layer 18 'relu3_1' 13 512 8 8 (425984)
  1562. I0818 13:35:11.972388 18451 layer_factory.hpp:172] Creating layer 'conv3_2' of type 'Convolution'
  1563. I0818 13:35:11.972391 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1564. I0818 13:35:11.972420 18451 net.cpp:202] Created Layer conv3_2 (19)
  1565. I0818 13:35:11.972426 18451 net.cpp:574] conv3_2 <- conv3_1
  1566. I0818 13:35:11.972432 18451 net.cpp:544] conv3_2 -> conv3_2
  1567. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1568. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1569. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1570. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1571. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1572. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1573. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1574. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1575. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1576. in math_function.cpp, line 730, tmp_value = 1048576
  1577. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1578. definition of type is half.
  1579. gspt r[i] = 0.00390625 (Q_frac=28
  1580. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1581. In blob.cpp, blob.Q_frac= 28
  1582. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1583. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1584. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1585. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1586. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1587. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1588. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1589. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1590. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1591. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1592. in math_function.cpp, line 1227, tmp_value = 0
  1593. definition of type is half [FILL ZERO].
  1594. gspt r[i] = 0 (Q_frac=22
  1595. In blob.cpp, blob.Q_frac= 22
  1596. I0818 13:35:12.539671 18451 net.cpp:262] Setting up conv3_2
  1597. I0818 13:35:12.539726 18451 net.cpp:269] TEST Top shape for layer 19 'conv3_2' 13 512 8 8 (425984)
  1598. I0818 13:35:12.539741 18451 layer_factory.hpp:172] Creating layer 'conv3_2_bn' of type 'BatchNorm'
  1599. I0818 13:35:12.539750 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1600. I0818 13:35:12.539788 18451 net.cpp:202] Created Layer conv3_2_bn (20)
  1601. I0818 13:35:12.539795 18451 net.cpp:574] conv3_2_bn <- conv3_2
  1602. I0818 13:35:12.539803 18451 net.cpp:529] conv3_2_bn -> conv3_2 (in-place)
  1603. I0818 13:35:12.540362 18451 net.cpp:262] Setting up conv3_2_bn
  1604. I0818 13:35:12.540371 18451 net.cpp:269] TEST Top shape for layer 20 'conv3_2_bn' 13 512 8 8 (425984)
  1605. I0818 13:35:12.540395 18451 layer_factory.hpp:172] Creating layer 'relu3_2' of type 'ReLU'
  1606. I0818 13:35:12.540416 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1607. I0818 13:35:12.540441 18451 net.cpp:202] Created Layer relu3_2 (21)
  1608. I0818 13:35:12.540446 18451 net.cpp:574] relu3_2 <- conv3_2
  1609. I0818 13:35:12.540450 18451 net.cpp:529] relu3_2 -> conv3_2 (in-place)
  1610. I0818 13:35:12.540459 18451 net.cpp:262] Setting up relu3_2
  1611. I0818 13:35:12.540467 18451 net.cpp:269] TEST Top shape for layer 21 'relu3_2' 13 512 8 8 (425984)
  1612. I0818 13:35:12.540475 18451 layer_factory.hpp:172] Creating layer 'pool3' of type 'Pooling'
  1613. I0818 13:35:12.540479 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1614. I0818 13:35:12.540493 18451 net.cpp:202] Created Layer pool3 (22)
  1615. I0818 13:35:12.540498 18451 net.cpp:574] pool3 <- conv3_2
  1616. I0818 13:35:12.540503 18451 net.cpp:544] pool3 -> pool3
  1617. I0818 13:35:12.540596 18451 net.cpp:262] Setting up pool3
  1618. I0818 13:35:12.540604 18451 net.cpp:269] TEST Top shape for layer 22 'pool3' 13 512 4 4 (106496)
  1619. I0818 13:35:12.540609 18451 layer_factory.hpp:172] Creating layer 'ip1' of type 'InnerProduct'
  1620. I0818 13:35:12.540617 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1621. I0818 13:35:12.540630 18451 net.cpp:202] Created Layer ip1 (23)
  1622. I0818 13:35:12.540635 18451 net.cpp:574] ip1 <- pool3
  1623. I0818 13:35:12.540642 18451 net.cpp:544] ip1 -> ip1
  1624. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1625. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1626. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1627. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1628. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1629. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1630. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1631. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1632. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1633. in math_function.cpp, line 730, tmp_value = 1048576
  1634. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1635. definition of type is half.
  1636. gspt r[i] = 0.00390625 (Q_frac=28
  1637. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1638. In blob.cpp, blob.Q_frac= 28
  1639. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1640. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1641. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1642. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1643. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1644. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1645. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1646. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1647. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1648. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1649. in math_function.cpp, line 1227, tmp_value = 0
  1650. definition of type is half [FILL ZERO].
  1651. gspt r[i] = 0 (Q_frac=22
  1652. In blob.cpp, blob.Q_frac= 22
  1653. I0818 13:35:14.681597 18451 net.cpp:262] Setting up ip1
  1654. I0818 13:35:14.681627 18451 net.cpp:269] TEST Top shape for layer 23 'ip1' 13 1024 (13312)
  1655. I0818 13:35:14.681670 18451 layer_factory.hpp:172] Creating layer 'relu4' of type 'ReLU'
  1656. I0818 13:35:14.681679 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1657. I0818 13:35:14.681690 18451 net.cpp:202] Created Layer relu4 (24)
  1658. I0818 13:35:14.681695 18451 net.cpp:574] relu4 <- ip1
  1659. I0818 13:35:14.681702 18451 net.cpp:529] relu4 -> ip1 (in-place)
  1660. I0818 13:35:14.681715 18451 net.cpp:262] Setting up relu4
  1661. I0818 13:35:14.681718 18451 net.cpp:269] TEST Top shape for layer 24 'relu4' 13 1024 (13312)
  1662. I0818 13:35:14.681722 18451 layer_factory.hpp:172] Creating layer 'drop7' of type 'Dropout'
  1663. I0818 13:35:14.681727 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1664. I0818 13:35:14.681814 18451 net.cpp:202] Created Layer drop7 (25)
  1665. I0818 13:35:14.681854 18451 net.cpp:574] drop7 <- ip1
  1666. I0818 13:35:14.681872 18451 net.cpp:529] drop7 -> ip1 (in-place)
  1667. I0818 13:35:14.692198 18451 net.cpp:262] Setting up drop7
  1668. I0818 13:35:14.692234 18451 net.cpp:269] TEST Top shape for layer 25 'drop7' 13 1024 (13312)
  1669. I0818 13:35:14.692247 18451 layer_factory.hpp:172] Creating layer 'ip2' of type 'InnerProduct'
  1670. I0818 13:35:14.692260 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1671. I0818 13:35:14.692286 18451 net.cpp:202] Created Layer ip2 (26)
  1672. I0818 13:35:14.692296 18451 net.cpp:574] ip2 <- ip1
  1673. I0818 13:35:14.692312 18451 net.cpp:544] ip2 -> ip2
  1674. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1675. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1676. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1677. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1678. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1679. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1680. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1681. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1682. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1683. in math_function.cpp, line 730, tmp_value = 1048576
  1684. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1685. definition of type is half.
  1686. gspt r[i] = 0.0078125 (Q_frac=27
  1687. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1688. In blob.cpp, blob.Q_frac= 27
  1689. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1690. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1691. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1692. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1693. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1694. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1695. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1696. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1697. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1698. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1699. in math_function.cpp, line 1227, tmp_value = 0
  1700. definition of type is half [FILL ZERO].
  1701. gspt r[i] = 0 (Q_frac=22
  1702. In blob.cpp, blob.Q_frac= 22
  1703. I0818 13:35:14.701611 18451 net.cpp:262] Setting up ip2
  1704. I0818 13:35:14.701633 18451 net.cpp:269] TEST Top shape for layer 26 'ip2' 13 10 (130)
  1705. I0818 13:35:14.701673 18451 layer_factory.hpp:172] Creating layer 'ip2_ip2_0_split' of type 'Split'
  1706. I0818 13:35:14.701683 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1707. I0818 13:35:14.701699 18451 net.cpp:202] Created Layer ip2_ip2_0_split (27)
  1708. I0818 13:35:14.701714 18451 net.cpp:574] ip2_ip2_0_split <- ip2
  1709. I0818 13:35:14.701725 18451 net.cpp:544] ip2_ip2_0_split -> ip2_ip2_0_split_0
  1710. I0818 13:35:14.701740 18451 net.cpp:544] ip2_ip2_0_split -> ip2_ip2_0_split_1
  1711. I0818 13:35:14.701844 18451 net.cpp:262] Setting up ip2_ip2_0_split
  1712. I0818 13:35:14.701858 18451 net.cpp:269] TEST Top shape for layer 27 'ip2_ip2_0_split' 13 10 (130)
  1713. I0818 13:35:14.701869 18451 net.cpp:269] TEST Top shape for layer 27 'ip2_ip2_0_split' 13 10 (130)
  1714. I0818 13:35:14.701879 18451 layer_factory.hpp:172] Creating layer 'accuracy' of type 'Accuracy'
  1715. I0818 13:35:14.701885 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1716. I0818 13:35:14.701925 18451 net.cpp:202] Created Layer accuracy (28)
  1717. I0818 13:35:14.701936 18451 net.cpp:574] accuracy <- ip2_ip2_0_split_0
  1718. I0818 13:35:14.701946 18451 net.cpp:574] accuracy <- label_cifar_1_split_0
  1719. I0818 13:35:14.701957 18451 net.cpp:544] accuracy -> accuracy
  1720. I0818 13:35:14.701983 18451 net.cpp:262] Setting up accuracy
  1721. I0818 13:35:14.701995 18451 net.cpp:269] TEST Top shape for layer 28 'accuracy' (1)
  1722. I0818 13:35:14.702003 18451 layer_factory.hpp:172] Creating layer 'loss' of type 'SoftmaxWithLoss'
  1723. I0818 13:35:14.702010 18451 layer_factory.hpp:184] Layer's types are Ftype:FLOAT16 Btype:FLOAT16 Fmath:FLOAT16 Bmath:FLOAT16
  1724. I0818 13:35:14.702033 18451 net.cpp:202] Created Layer loss (29)
  1725. I0818 13:35:14.702042 18451 net.cpp:574] loss <- ip2_ip2_0_split_1
  1726. I0818 13:35:14.702051 18451 net.cpp:574] loss <- label_cifar_1_split_1
  1727. I0818 13:35:14.702059 18451 net.cpp:544] loss -> loss
  1728. I0818 13:35:14.702523 18451 net.cpp:262] Setting up loss
  1729. I0818 13:35:14.702540 18451 net.cpp:269] TEST Top shape for layer 29 'loss' (1)
  1730. I0818 13:35:14.702548 18451 net.cpp:273] with loss weight 1
  1731. I0818 13:35:14.702560 18451 net.cpp:338] loss needs backward computation.
  1732. I0818 13:35:14.702570 18451 net.cpp:340] accuracy does not need backward computation.
  1733. I0818 13:35:14.702582 18451 net.cpp:338] ip2_ip2_0_split needs backward computation.
  1734. I0818 13:35:14.702589 18451 net.cpp:338] ip2 needs backward computation.
  1735. I0818 13:35:14.702595 18451 net.cpp:338] drop7 needs backward computation.
  1736. I0818 13:35:14.702631 18451 net.cpp:338] relu4 needs backward computation.
  1737. I0818 13:35:14.702641 18451 net.cpp:338] ip1 needs backward computation.
  1738. I0818 13:35:14.702658 18451 net.cpp:338] pool3 needs backward computation.
  1739. I0818 13:35:14.702667 18451 net.cpp:338] relu3_2 needs backward computation.
  1740. I0818 13:35:14.702673 18451 net.cpp:338] conv3_2_bn needs backward computation.
  1741. I0818 13:35:14.702682 18451 net.cpp:338] conv3_2 needs backward computation.
  1742. I0818 13:35:14.702689 18451 net.cpp:338] relu3_1 needs backward computation.
  1743. I0818 13:35:14.702695 18451 net.cpp:338] conv3_1_bn needs backward computation.
  1744. I0818 13:35:14.702702 18451 net.cpp:338] conv3_1 needs backward computation.
  1745. I0818 13:35:14.702710 18451 net.cpp:338] pool2 needs backward computation.
  1746. I0818 13:35:14.702718 18451 net.cpp:338] relu2_2 needs backward computation.
  1747. I0818 13:35:14.702723 18451 net.cpp:338] conv2_2_bn needs backward computation.
  1748. I0818 13:35:14.702733 18451 net.cpp:338] conv2_2 needs backward computation.
  1749. I0818 13:35:14.702746 18451 net.cpp:338] relu2_1 needs backward computation.
  1750. I0818 13:35:14.702754 18451 net.cpp:338] conv2_1_bn needs backward computation.
  1751. I0818 13:35:14.702764 18451 net.cpp:338] conv2_1 needs backward computation.
  1752. I0818 13:35:14.702772 18451 net.cpp:338] pool1 needs backward computation.
  1753. I0818 13:35:14.702783 18451 net.cpp:338] relu1_2 needs backward computation.
  1754. I0818 13:35:14.702790 18451 net.cpp:338] conv1_2_bn needs backward computation.
  1755. I0818 13:35:14.702800 18451 net.cpp:338] conv1_2 needs backward computation.
  1756. I0818 13:35:14.702808 18451 net.cpp:338] relu1_1 needs backward computation.
  1757. I0818 13:35:14.702818 18451 net.cpp:338] conv1_1_bn needs backward computation.
  1758. I0818 13:35:14.702824 18451 net.cpp:338] conv1_1 needs backward computation.
  1759. I0818 13:35:14.702832 18451 net.cpp:340] label_cifar_1_split does not need backward computation.
  1760. I0818 13:35:14.702841 18451 net.cpp:340] cifar does not need backward computation.
  1761. I0818 13:35:14.702850 18451 net.cpp:382] This network produces output accuracy
  1762. I0818 13:35:14.702857 18451 net.cpp:382] This network produces output loss
  1763. I0818 13:35:14.702935 18451 net.cpp:405] Top memory (TEST) required for data: 37434216 diff: 37514116
  1764. I0818 13:35:14.702944 18451 net.cpp:408] Bottom memory (TEST) required for data: 37434208 diff: 37514108
  1765. I0818 13:35:14.702952 18451 net.cpp:411] Shared (in-place) memory (TEST) by data: 23908352 diff: 23908352
  1766. I0818 13:35:14.702960 18451 net.cpp:414] Parameters memory (TEST) required for data: 25966916 diff: 25966916
  1767. I0818 13:35:14.702965 18451 net.cpp:417] Parameters shared memory (TEST) by data: 0 diff: 0
  1768. I0818 13:35:14.702971 18451 net.cpp:423] Network initialization done.
  1769. I0818 13:35:14.703248 18451 solver.cpp:55] Solver scaffolding done.
  1770. I0818 13:35:14.794719 18451 parallel.cpp:95] Starting sync 0 (of total 8), {0.8}
  1771. I0818 13:35:14.794766 18451 parallel.cpp:169] [0 - 0] P2PSync adding callback
  1772. I0818 13:35:14.794772 18451 parallel.cpp:95] Starting sync 1 (of total 8), {1.8}
  1773. I0818 13:35:14.794780 18451 parallel.cpp:169] [1 - 1] P2PSync adding callback
  1774. I0818 13:35:14.794783 18451 parallel.cpp:95] Starting sync 2 (of total 8), {2.8}
  1775. I0818 13:35:14.794790 18451 parallel.cpp:169] [2 - 2] P2PSync adding callback
  1776. I0818 13:35:14.794792 18451 parallel.cpp:95] Starting sync 3 (of total 8), {3.8}
  1777. I0818 13:35:14.794798 18451 parallel.cpp:169] [3 - 3] P2PSync adding callback
  1778. I0818 13:35:14.794802 18451 parallel.cpp:95] Starting sync 4 (of total 8), {4.8}
  1779. I0818 13:35:14.794807 18451 parallel.cpp:169] [4 - 4] P2PSync adding callback
  1780. I0818 13:35:14.794811 18451 parallel.cpp:95] Starting sync 5 (of total 8), {5.8}
  1781. I0818 13:35:14.794817 18451 parallel.cpp:169] [5 - 5] P2PSync adding callback
  1782. I0818 13:35:14.794821 18451 parallel.cpp:95] Starting sync 6 (of total 8), {6.8}
  1783. I0818 13:35:14.794826 18451 parallel.cpp:169] [6 - 6] P2PSync adding callback
  1784. I0818 13:35:14.794828 18451 parallel.cpp:95] Starting sync 7 (of total 8), {7.8}
  1785. I0818 13:35:14.794834 18451 parallel.cpp:169] [7 - 7] P2PSync adding callback
  1786. I0818 13:35:14.794860 18451 parallel.cpp:101] Starting Optimization
  1787. I0818 13:35:14.794865 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 0
  1788. I0818 13:35:14.794921 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 1
  1789. I0818 13:35:14.794972 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 2
  1790. I0818 13:35:14.795083 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 3
  1791. I0818 13:35:14.795147 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 4
  1792. I0818 13:35:14.795286 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 5
  1793. I0818 13:35:14.795805 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 6
  1794. I0818 13:35:14.795949 18451 internal_thread.cpp:18] Starting 1 internal thread(s) on device 7
  1795. I0818 13:35:14.826941 18523 common.cpp:533] NVML initialized, thread 18523
  1796. I0818 13:35:14.944767 18523 common.cpp:555] {0} NVML succeeded to set CPU affinity
  1797. I0818 13:35:15.144954 18524 common.cpp:555] {1} NVML succeeded to set CPU affinity
  1798. I0818 13:35:15.483872 18525 common.cpp:555] {2} NVML succeeded to set CPU affinity
  1799. I0818 13:35:15.860432 18526 common.cpp:555] {3} NVML succeeded to set CPU affinity
  1800. I0818 13:35:16.245779 18528 common.cpp:555] {5} NVML succeeded to set CPU affinity
  1801. I0818 13:35:16.650892 18527 common.cpp:555] {4} NVML succeeded to set CPU affinity
  1802. I0818 13:35:17.082074 18529 common.cpp:555] {6} NVML succeeded to set CPU affinity
  1803. I0818 13:35:17.485327 18525 solver.cpp:40] Solver data type: FLOAT16
  1804. I0818 13:35:17.485350 18526 solver.cpp:40] Solver data type: FLOAT16
  1805. I0818 13:35:17.485405 18528 solver.cpp:40] Solver data type: FLOAT16
  1806. I0818 13:35:17.485476 18527 solver.cpp:40] Solver data type: FLOAT16
  1807. I0818 13:35:17.485510 18529 solver.cpp:40] Solver data type: FLOAT16
  1808. I0818 13:35:17.485327 18524 solver.cpp:40] Solver data type: FLOAT16
  1809. W0818 13:35:17.489166 18525 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  1810. I0818 13:35:17.489627 18525 internal_thread.cpp:18] Starting 1 internal thread(s) on device 2
  1811. I0818 13:35:17.494256 18530 common.cpp:555] {7} NVML succeeded to set CPU affinity
  1812. I0818 13:35:17.496601 18530 solver.cpp:40] Solver data type: FLOAT16
  1813. W0818 13:35:17.499847 18528 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  1814. I0818 13:35:17.500329 18528 internal_thread.cpp:18] Starting 1 internal thread(s) on device 5
  1815. W0818 13:35:17.515105 18526 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  1816. I0818 13:35:17.515568 18526 internal_thread.cpp:18] Starting 1 internal thread(s) on device 3
  1817. I0818 13:35:17.521049 18525 data_reader.cpp:59] Sample Data Reader threads: 1, out queues: 1, depth: 13
  1818. I0818 13:35:17.521095 18525 internal_thread.cpp:18] Starting 1 internal thread(s) on device 2
  1819. I0818 13:35:17.524359 18548 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  1820. W0818 13:35:17.531006 18527 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  1821. I0818 13:35:17.531407 18527 internal_thread.cpp:18] Starting 1 internal thread(s) on device 4
  1822. I0818 13:35:17.545449 18528 data_reader.cpp:59] Sample Data Reader threads: 1, out queues: 1, depth: 13
  1823. I0818 13:35:17.545488 18528 internal_thread.cpp:18] Starting 1 internal thread(s) on device 5
  1824. I0818 13:35:17.547411 18550 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  1825. W0818 13:35:17.547865 18529 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  1826. I0818 13:35:17.548334 18529 internal_thread.cpp:18] Starting 1 internal thread(s) on device 6
  1827. I0818 13:35:17.548405 18526 data_reader.cpp:59] Sample Data Reader threads: 1, out queues: 1, depth: 13
  1828. I0818 13:35:17.548446 18526 internal_thread.cpp:18] Starting 1 internal thread(s) on device 3
  1829. I0818 13:35:17.553680 18552 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  1830. I0818 13:35:17.558228 18525 data_layer.cpp:200] [n0.d2.r2] Output data size: 13, 3, 28, 28
  1831. W0818 13:35:17.562413 18524 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  1832. I0818 13:35:17.562922 18524 internal_thread.cpp:18] Starting 1 internal thread(s) on device 1
  1833. I0818 13:35:17.568377 18527 data_reader.cpp:59] Sample Data Reader threads: 1, out queues: 1, depth: 13
  1834. I0818 13:35:17.568423 18527 internal_thread.cpp:18] Starting 1 internal thread(s) on device 4
  1835. I0818 13:35:17.574710 18554 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  1836. I0818 13:35:17.583221 18525 internal_thread.cpp:18] Starting 1 internal thread(s) on device 2
  1837. I0818 13:35:17.589362 18526 data_layer.cpp:200] [n0.d3.r3] Output data size: 13, 3, 28, 28
  1838. I0818 13:35:17.590121 18524 data_reader.cpp:59] Sample Data Reader threads: 1, out queues: 1, depth: 13
  1839. I0818 13:35:17.590159 18524 internal_thread.cpp:18] Starting 1 internal thread(s) on device 1
  1840. I0818 13:35:17.590986 18528 data_layer.cpp:200] [n0.d5.r5] Output data size: 13, 3, 28, 28
  1841. I0818 13:35:17.592882 18527 data_layer.cpp:200] [n0.d4.r4] Output data size: 13, 3, 28, 28
  1842. I0818 13:35:17.594213 18556 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  1843. I0818 13:35:17.594269 18529 data_reader.cpp:59] Sample Data Reader threads: 1, out queues: 1, depth: 13
  1844. I0818 13:35:17.594312 18529 internal_thread.cpp:18] Starting 1 internal thread(s) on device 6
  1845. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1846. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1847. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1848. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1849. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1850. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1851. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1852. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1853. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1854. in math_function.cpp, line 730, tmp_value = 1048576
  1855. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1856. definition of type is half.
  1857. gspt r[i] = 0.0625 (Q_frac=24
  1858. I0818 13:35:17.596905 18524 data_layer.cpp:200] [n0.d1.r1] Output data size: 13, 3, 28, 28
  1859. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1860. In blob.cpp, blob.Q_frac= 24
  1861. I0818 13:35:17.598275 18557 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  1862. W0818 13:35:17.598385 18530 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  1863. I0818 13:35:17.598388 18526 internal_thread.cpp:18] Starting 1 internal thread(s) on device 3
  1864. I0818 13:35:17.600096 18527 internal_thread.cpp:18] Starting 1 internal thread(s) on device 4
  1865. I0818 13:35:17.600111 18528 internal_thread.cpp:18] Starting 1 internal thread(s) on device 5
  1866. I0818 13:35:17.600970 18530 internal_thread.cpp:18] Starting 1 internal thread(s) on device 7
  1867. I0818 13:35:17.605684 18524 internal_thread.cpp:18] Starting 1 internal thread(s) on device 1
  1868. I0818 13:35:17.608249 18529 data_layer.cpp:200] [n0.d6.r6] Output data size: 13, 3, 28, 28
  1869. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1870. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1871. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1872. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1873. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1874. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1875. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1876. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1877. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1878. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1879. in math_function.cpp, line 1227, tmp_value = 0
  1880. definition of type is half [FILL ZERO].
  1881. gspt r[i] = 0 (Q_frac=22
  1882. In blob.cpp, blob.Q_frac= 22
  1883. I0818 13:35:17.624114 18529 internal_thread.cpp:18] Starting 1 internal thread(s) on device 6
  1884. I0818 13:35:17.627607 18530 data_reader.cpp:59] Sample Data Reader threads: 1, out queues: 1, depth: 13
  1885. I0818 13:35:17.627657 18530 internal_thread.cpp:18] Starting 1 internal thread(s) on device 7
  1886. I0818 13:35:17.629189 18564 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  1887. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1888. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1889. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1890. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1891. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1892. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1893. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1894. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1895. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1896. in math_function.cpp, line 730, tmp_value = 1048576
  1897. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1898. definition of type is half.
  1899. gspt r[i] = 0.0625 (Q_frac=24
  1900. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1901. In blob.cpp, blob.Q_frac= 24
  1902. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1903. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1904. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1905. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1906. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1907. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1908. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1909. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1910. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1911. in math_function.cpp, line 730, tmp_value = 1048576
  1912. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1913. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1914. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1915. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1916. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1917. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1918. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1919. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1920. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1921. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1922. in math_function.cpp, line 730, tmp_value = 1048576
  1923. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1924. definition of type is half.
  1925. gspt r[i] = 0.0625 (Q_frac=24
  1926. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1927. In blob.cpp, blob.Q_frac= 24
  1928. definition of type is half.
  1929. gspt r[i] = 0.0625 (Q_frac=24
  1930. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1931. In blob.cpp, blob.Q_frac= 24
  1932. I0818 13:35:17.696408 18530 data_layer.cpp:200] [n0.d7.r7] Output data size: 13, 3, 28, 28
  1933. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1934. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1935. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1936. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1937. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1938. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1939. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1940. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1941. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1942. in math_function.cpp, line 730, tmp_value = 1048576
  1943. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1944. definition of type is half.
  1945. gspt r[i] = 0.0625 (Q_frac=24
  1946. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1947. In blob.cpp, blob.Q_frac= 24
  1948. I0818 13:35:17.709332 18530 internal_thread.cpp:18] Starting 1 internal thread(s) on device 7
  1949. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1950. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1951. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1952. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1953. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1954. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1955. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1956. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1957. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1958. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1959. in math_function.cpp, line 1227, tmp_value = 0
  1960. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  1961. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  1962. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  1963. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  1964. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  1965. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  1966. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  1967. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  1968. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  1969. in math_function.cpp, line 730, tmp_value = 1048576
  1970. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  1971. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1972. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1973. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1974. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1975. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1976. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1977. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1978. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  1979. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  1980. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  1981. in math_function.cpp, line 1227, tmp_value = 0
  1982. definition of type is half [FILL ZERO].
  1983. gspt r[i] = 0 (Q_frac=22
  1984. In blob.cpp, blob.Q_frac= 22
  1985. definition of type is half.
  1986. gspt r[i] = 0.0625 (Q_frac=24
  1987. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  1988. In blob.cpp, blob.Q_frac= 24
  1989. definition of type is half [FILL ZERO].
  1990. gspt r[i] = 0 (Q_frac=22
  1991. In blob.cpp, blob.Q_frac= 22
  1992. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  1993. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  1994. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  1995. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  1996. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  1997. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  1998. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  1999. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2000. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2001. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2002. in math_function.cpp, line 1227, tmp_value = 0
  2003. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2004. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2005. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2006. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2007. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2008. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2009. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2010. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2011. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2012. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2013. in math_function.cpp, line 1227, tmp_value = 0
  2014. definition of type is half [FILL ZERO].
  2015. gspt r[i] = 0 (Q_frac=22
  2016. In blob.cpp, blob.Q_frac= 22
  2017. definition of type is half [FILL ZERO].
  2018. gspt r[i] = 0 (Q_frac=22
  2019. In blob.cpp, blob.Q_frac= 22
  2020. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2021. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2022. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2023. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2024. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2025. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2026. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2027. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2028. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2029. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2030. in math_function.cpp, line 1227, tmp_value = 0
  2031. definition of type is half [FILL ZERO].
  2032. gspt r[i] = 0 (Q_frac=22
  2033. In blob.cpp, blob.Q_frac= 22
  2034. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2035. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2036. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2037. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2038. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2039. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2040. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2041. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2042. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2043. in math_function.cpp, line 730, tmp_value = 1048576
  2044. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  2045. definition of type is half.
  2046. gspt r[i] = 0.0625 (Q_frac=24
  2047. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2048. In blob.cpp, blob.Q_frac= 24
  2049. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2050. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2051. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2052. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2053. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2054. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2055. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2056. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2057. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2058. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2059. in math_function.cpp, line 1227, tmp_value = 0
  2060. definition of type is half [FILL ZERO].
  2061. gspt r[i] = 0 (Q_frac=22
  2062. In blob.cpp, blob.Q_frac= 22
  2063. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2064. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2065. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2066. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2067. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2068. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2069. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2070. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2071. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2072. in math_function.cpp, line 730, tmp_value = 1048576
  2073. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  2074. definition of type is half.
  2075. gspt r[i] = 0.0078125 (Q_frac=27
  2076. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2077. In blob.cpp, blob.Q_frac= 27
  2078. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2079. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2080. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2081. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2082. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2083. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2084. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2085. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2086. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2087. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2088. in math_function.cpp, line 1227, tmp_value = 0
  2089. definition of type is half [FILL ZERO].
  2090. gspt r[i] = 0 (Q_frac=22
  2091. In blob.cpp, blob.Q_frac= 22
  2092. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2093. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2094. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2095. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2096. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2097. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2098. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2099. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2100. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2101. in math_function.cpp, line 730, tmp_value = 1048576
  2102. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  2103. definition of type is half.
  2104. gspt r[i] = 0.0078125 (Q_frac=27
  2105. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2106. In blob.cpp, blob.Q_frac= 27
  2107. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2108. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2109. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2110. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2111. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2112. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2113. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2114. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2115. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2116. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2117. in math_function.cpp, line 1227, tmp_value = 0
  2118. definition of type is half [FILL ZERO].
  2119. gspt r[i] = 0 (Q_frac=22
  2120. In blob.cpp, blob.Q_frac= 22
  2121. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2122. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2123. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2124. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2125. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2126. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2127. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2128. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2129. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2130. in math_function.cpp, line 730, tmp_value = 1048576
  2131. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  2132. definition of type is half.
  2133. gspt r[i] = 0.0078125 (Q_frac=27
  2134. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2135. In blob.cpp, blob.Q_frac= 27
  2136. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2137. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2138. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2139. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2140. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2141. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2142. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2143. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2144. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2145. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2146. in math_function.cpp, line 1227, tmp_value = 0
  2147. definition of type is half [FILL ZERO].
  2148. gspt r[i] = 0 (Q_frac=22
  2149. In blob.cpp, blob.Q_frac= 22
  2150. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2151. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2152. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2153. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2154. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2155. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2156. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2157. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2158. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2159. in math_function.cpp, line 730, tmp_value = 1048576
  2160. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  2161. definition of type is half.
  2162. gspt r[i] = 0.0078125 (Q_frac=27
  2163. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2164. In blob.cpp, blob.Q_frac= 27
  2165. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2166. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2167. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2168. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2169. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2170. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2171. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2172. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2173. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2174. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2175. in math_function.cpp, line 1227, tmp_value = 0
  2176. definition of type is half [FILL ZERO].
  2177. gspt r[i] = 0 (Q_frac=22
  2178. In blob.cpp, blob.Q_frac= 22
  2179. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2180. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2181. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2182. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2183. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2184. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2185. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2186. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2187. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2188. in math_function.cpp, line 730, tmp_value = 1048576
  2189. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  2190. definition of type is half.
  2191. gspt r[i] = 0.0078125 (Q_frac=27
  2192. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2193. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2194. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2195. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2196. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2197. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  2198. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  2199. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  2200. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  2201. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2202. in math_function.cpp, line 730, tmp_value = -1179648
  2203. in math_function.cpp line 656, gspt_info_1[i]= 9699401, gspt_info_2[i]= 0
  2204. definition of type is half.
  2205. gspt r[i] = -0.00878906 (Q_frac=27
  2206. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2207. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  2208. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  2209. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  2210. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2211. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  2212. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2213. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2214. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  2215. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2216. in math_function.cpp, line 730, tmp_value = 524288
  2217. in math_function.cpp line 656, gspt_info_1[i]= 4527444, gspt_info_2[i]= 0
  2218. definition of type is half.
  2219. gspt r[i] = 0.00390625 (Q_frac=27
  2220. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2221. In blob.cpp, blob.Q_frac= 27
  2222. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2223. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2224. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2225. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2226. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2227. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2228. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2229. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2230. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2231. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2232. in math_function.cpp, line 1227, tmp_value = 0
  2233. definition of type is half [FILL ZERO].
  2234. gspt r[i] = 0 (Q_frac=22
  2235. In blob.cpp, blob.Q_frac= 22
  2236. In blob.cpp, blob.Q_frac= 27
  2237. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2238. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  2239. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2240. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2241. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  2242. in math_function.cpp line 565, grp_num= 5, rand_index= 6
  2243. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  2244. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  2245. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2246. in math_function.cpp, line 730, tmp_value = 1966080
  2247. in math_function.cpp line 656, gspt_info_1[i]= 8917902, gspt_info_2[i]= 0
  2248. definition of type is half.
  2249. gspt r[i] = 0.0146484 (Q_frac=27
  2250. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2251. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2252. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2253. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2254. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2255. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2256. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2257. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2258. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2259. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2260. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2261. in math_function.cpp, line 1227, tmp_value = 0
  2262. definition of type is half [FILL ZERO].
  2263. gspt r[i] = 0 (Q_frac=22
  2264. In blob.cpp, blob.Q_frac= 22
  2265. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2266. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  2267. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  2268. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  2269. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  2270. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2271. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  2272. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  2273. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2274. in math_function.cpp, line 730, tmp_value = 786432
  2275. in math_function.cpp line 656, gspt_info_1[i]= 8860836, gspt_info_2[i]= 0
  2276. definition of type is half.
  2277. gspt r[i] = 0.00585938 (Q_frac=27
  2278. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2279. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  2280. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  2281. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  2282. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2283. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  2284. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2285. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  2286. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  2287. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2288. in math_function.cpp, line 730, tmp_value = 1966080
  2289. in math_function.cpp line 656, gspt_info_1[i]= 12194958, gspt_info_2[i]= 0
  2290. definition of type is half.
  2291. gspt r[i] = 0.0146484 (Q_frac=27
  2292. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2293. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2294. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  2295. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  2296. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2297. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  2298. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  2299. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  2300. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  2301. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2302. in math_function.cpp, line 730, tmp_value = -1835008
  2303. in math_function.cpp line 656, gspt_info_1[i]= 623904, gspt_info_2[i]= 0
  2304. definition of type is half.
  2305. gspt r[i] = -0.0136719 (Q_frac=27
  2306. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2307. In blob.cpp, blob.Q_frac= 27
  2308. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2309. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2310. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2311. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2312. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2313. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2314. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2315. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2316. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2317. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2318. in math_function.cpp, line 1227, tmp_value = 0
  2319. definition of type is half [FILL ZERO].
  2320. gspt r[i] = 0 (Q_frac=22
  2321. In blob.cpp, blob.Q_frac= 22
  2322. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2323. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  2324. in math_function.cpp line 565, grp_num= 2, rand_index= 1
  2325. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2326. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  2327. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2328. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  2329. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  2330. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2331. in math_function.cpp, line 730, tmp_value = -262144
  2332. in math_function.cpp line 656, gspt_info_1[i]= 1081347, gspt_info_2[i]= 0
  2333. definition of type is half.
  2334. gspt r[i] = -0.00195312 (Q_frac=27
  2335. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2336. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2337. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  2338. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  2339. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  2340. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  2341. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2342. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  2343. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  2344. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2345. in math_function.cpp, line 730, tmp_value = -1310720
  2346. in math_function.cpp line 656, gspt_info_1[i]= 76417, gspt_info_2[i]= 0
  2347. definition of type is half.
  2348. gspt r[i] = -0.00976562 (Q_frac=27
  2349. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2350. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2351. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  2352. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  2353. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  2354. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  2355. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2356. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  2357. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  2358. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2359. in math_function.cpp, line 730, tmp_value = -917504
  2360. in math_function.cpp line 656, gspt_info_1[i]= 1225753, gspt_info_2[i]= 0
  2361. definition of type is half.
  2362. gspt r[i] = -0.00683594 (Q_frac=27
  2363. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2364. In blob.cpp, blob.Q_frac= 27
  2365. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2366. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2367. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2368. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2369. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2370. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2371. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2372. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2373. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2374. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2375. in math_function.cpp, line 1227, tmp_value = 0
  2376. definition of type is half [FILL ZERO].
  2377. gspt r[i] = 0 (Q_frac=22
  2378. In blob.cpp, blob.Q_frac= 22
  2379. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2380. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  2381. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  2382. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  2383. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  2384. in math_function.cpp line 565, grp_num= 5, rand_index= 6
  2385. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  2386. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2387. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2388. in math_function.cpp, line 730, tmp_value = 1179648
  2389. in math_function.cpp line 656, gspt_info_1[i]= 1222557, gspt_info_2[i]= 0
  2390. definition of type is half.
  2391. gspt r[i] = 0.00878906 (Q_frac=27
  2392. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2393. In blob.cpp, blob.Q_frac= 27
  2394. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2395. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2396. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2397. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2398. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2399. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2400. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2401. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2402. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2403. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2404. in math_function.cpp, line 1227, tmp_value = 0
  2405. definition of type is half [FILL ZERO].
  2406. gspt r[i] = 0 (Q_frac=22
  2407. In blob.cpp, blob.Q_frac= 22
  2408. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2409. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  2410. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  2411. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2412. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2413. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2414. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  2415. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  2416. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2417. in math_function.cpp, line 730, tmp_value = 1835008
  2418. in math_function.cpp line 656, gspt_info_1[i]= 1252486, gspt_info_2[i]= 0
  2419. definition of type is half.
  2420. gspt r[i] = 0.0136719 (Q_frac=27
  2421. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2422. In blob.cpp, blob.Q_frac= 27
  2423. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2424. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2425. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2426. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2427. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2428. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2429. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2430. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2431. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2432. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2433. in math_function.cpp, line 1227, tmp_value = 0
  2434. definition of type is half [FILL ZERO].
  2435. gspt r[i] = 0 (Q_frac=22
  2436. In blob.cpp, blob.Q_frac= 22
  2437. in math_function.cpp line 565, grp_num= 0, rand_index= 3
  2438. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  2439. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  2440. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  2441. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2442. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2443. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  2444. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  2445. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2446. in math_function.cpp, line 730, tmp_value = 262144
  2447. in math_function.cpp line 656, gspt_info_1[i]= 7220259, gspt_info_2[i]= 0
  2448. definition of type is half.
  2449. gspt r[i] = 0.00195312 (Q_frac=27
  2450. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2451. In blob.cpp, blob.Q_frac= 27
  2452. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2453. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2454. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2455. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2456. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2457. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2458. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2459. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2460. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2461. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2462. in math_function.cpp, line 1227, tmp_value = 0
  2463. definition of type is half [FILL ZERO].
  2464. gspt r[i] = 0 (Q_frac=22
  2465. In blob.cpp, blob.Q_frac= 22
  2466. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2467. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  2468. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  2469. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2470. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  2471. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2472. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  2473. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  2474. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2475. in math_function.cpp, line 730, tmp_value = 262144
  2476. in math_function.cpp line 656, gspt_info_1[i]= 8519843, gspt_info_2[i]= 0
  2477. In blob.cpp, blob.Q_frac= 27
  2478. definition of type is half.
  2479. gspt r[i] = 0.000976562 (Q_frac=28
  2480. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2481. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2482. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2483. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2484. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2485. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2486. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2487. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2488. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2489. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2490. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2491. in math_function.cpp, line 1227, tmp_value = 0
  2492. definition of type is half [FILL ZERO].
  2493. gspt r[i] = 0 (Q_frac=22
  2494. In blob.cpp, blob.Q_frac= 22
  2495. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2496. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  2497. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  2498. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  2499. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  2500. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2501. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  2502. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  2503. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2504. in math_function.cpp, line 730, tmp_value = -917504
  2505. in math_function.cpp line 656, gspt_info_1[i]= 178009, gspt_info_2[i]= 0
  2506. definition of type is half.
  2507. gspt r[i] = -0.00683594 (Q_frac=27
  2508. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2509. In blob.cpp, blob.Q_frac= 27
  2510. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2511. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2512. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2513. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2514. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2515. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2516. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2517. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2518. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2519. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2520. in math_function.cpp, line 1227, tmp_value = 0
  2521. definition of type is half [FILL ZERO].
  2522. gspt r[i] = 0 (Q_frac=22
  2523. In blob.cpp, blob.Q_frac= 22
  2524. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2525. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  2526. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  2527. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  2528. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  2529. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2530. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2531. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  2532. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2533. in math_function.cpp, line 730, tmp_value = -524288
  2534. in math_function.cpp line 656, gspt_info_1[i]= 148818, gspt_info_2[i]= 0
  2535. definition of type is half.
  2536. gspt r[i] = -0.00390625 (Q_frac=27
  2537. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2538. In blob.cpp, blob.Q_frac= 27
  2539. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2540. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2541. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2542. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2543. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2544. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2545. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2546. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2547. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2548. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2549. in math_function.cpp, line 1227, tmp_value = 0
  2550. definition of type is half [FILL ZERO].
  2551. gspt r[i] = 0 (Q_frac=22
  2552. In blob.cpp, blob.Q_frac= 22
  2553. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  2554. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  2555. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  2556. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2557. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  2558. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2559. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  2560. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  2561. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2562. in math_function.cpp, line 730, tmp_value = -786432
  2563. in math_function.cpp line 656, gspt_info_1[i]= 5935457, gspt_info_2[i]= 0
  2564. definition of type is half.
  2565. gspt r[i] = -0.00585938 (Q_frac=27
  2566. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2567. In blob.cpp, blob.Q_frac= 27
  2568. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2569. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2570. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2571. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2572. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2573. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2574. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2575. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2576. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2577. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2578. in math_function.cpp, line 1227, tmp_value = 0
  2579. definition of type is half [FILL ZERO].
  2580. gspt r[i] = 0 (Q_frac=22
  2581. In blob.cpp, blob.Q_frac= 22
  2582. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  2583. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  2584. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  2585. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2586. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  2587. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2588. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  2589. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  2590. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2591. in math_function.cpp, line 730, tmp_value = -1179648
  2592. in math_function.cpp line 656, gspt_info_1[i]= 12980553, gspt_info_2[i]= 0
  2593. definition of type is half.
  2594. gspt r[i] = -0.00878906 (Q_frac=27
  2595. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2596. In blob.cpp, blob.Q_frac= 27
  2597. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2598. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2599. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2600. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2601. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2602. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2603. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2604. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2605. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2606. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2607. in math_function.cpp, line 1227, tmp_value = 0
  2608. definition of type is half [FILL ZERO].
  2609. gspt r[i] = 0 (Q_frac=22
  2610. In blob.cpp, blob.Q_frac= 22
  2611. in math_function.cpp line 565, grp_num= 0, rand_index= 1
  2612. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  2613. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  2614. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  2615. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  2616. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  2617. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  2618. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  2619. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2620. in math_function.cpp, line 730, tmp_value = -131072
  2621. in math_function.cpp line 656, gspt_info_1[i]= 2800331, gspt_info_2[i]= 0
  2622. definition of type is half.
  2623. gspt r[i] = -0.000976562 (Q_frac=27
  2624. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2625. In blob.cpp, blob.Q_frac= 27
  2626. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2627. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2628. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2629. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2630. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2631. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2632. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2633. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2634. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2635. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2636. in math_function.cpp, line 1227, tmp_value = 0
  2637. definition of type is half [FILL ZERO].
  2638. gspt r[i] = 0 (Q_frac=22
  2639. In blob.cpp, blob.Q_frac= 22
  2640. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  2641. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  2642. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  2643. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2644. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  2645. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2646. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  2647. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  2648. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2649. in math_function.cpp, line 730, tmp_value = -1966080
  2650. in math_function.cpp line 656, gspt_info_1[i]= 4823704, gspt_info_2[i]= 0
  2651. definition of type is half.
  2652. gspt r[i] = -0.0146484 (Q_frac=27
  2653. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2654. In blob.cpp, blob.Q_frac= 27
  2655. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2656. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2657. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2658. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2659. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2660. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2661. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2662. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2663. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2664. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2665. in math_function.cpp, line 1227, tmp_value = 0
  2666. definition of type is half [FILL ZERO].
  2667. gspt r[i] = 0 (Q_frac=22
  2668. In blob.cpp, blob.Q_frac= 22
  2669. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  2670. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2671. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  2672. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  2673. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  2674. in math_function.cpp line 565, grp_num= 5, rand_index= 6
  2675. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  2676. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2677. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2678. in math_function.cpp, line 730, tmp_value = 917504
  2679. in math_function.cpp line 656, gspt_info_1[i]= 11904397, gspt_info_2[i]= 0
  2680. definition of type is half.
  2681. gspt r[i] = 0.00683594 (Q_frac=27
  2682. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2683. In blob.cpp, blob.Q_frac= 27
  2684. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2685. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2686. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2687. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2688. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2689. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2690. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2691. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2692. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2693. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2694. in math_function.cpp, line 1227, tmp_value = 0
  2695. definition of type is half [FILL ZERO].
  2696. gspt r[i] = 0 (Q_frac=22
  2697. In blob.cpp, blob.Q_frac= 22
  2698. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  2699. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  2700. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  2701. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  2702. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  2703. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2704. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  2705. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  2706. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2707. in math_function.cpp, line 730, tmp_value = -262144
  2708. in math_function.cpp line 656, gspt_info_1[i]= 13316931, gspt_info_2[i]= 0
  2709. definition of type is half.
  2710. gspt r[i] = -0.00195312 (Q_frac=27
  2711. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2712. In blob.cpp, blob.Q_frac= 27
  2713. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2714. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2715. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2716. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2717. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2718. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2719. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2720. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2721. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2722. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2723. in math_function.cpp, line 1227, tmp_value = 0
  2724. definition of type is half [FILL ZERO].
  2725. gspt r[i] = 0 (Q_frac=22
  2726. In blob.cpp, blob.Q_frac= 22
  2727. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  2728. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  2729. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  2730. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  2731. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  2732. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  2733. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  2734. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  2735. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2736. in math_function.cpp, line 730, tmp_value = -1835008
  2737. in math_function.cpp line 656, gspt_info_1[i]= 11381472, gspt_info_2[i]= 0
  2738. definition of type is half.
  2739. gspt r[i] = -0.0136719 (Q_frac=27
  2740. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2741. In blob.cpp, blob.Q_frac= 27
  2742. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2743. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2744. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2745. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2746. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2747. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2748. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2749. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2750. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2751. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2752. in math_function.cpp, line 1227, tmp_value = 0
  2753. definition of type is half [FILL ZERO].
  2754. gspt r[i] = 0 (Q_frac=22
  2755. In blob.cpp, blob.Q_frac= 22
  2756. in math_function.cpp line 565, grp_num= 0, rand_index= 3
  2757. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  2758. in math_function.cpp line 565, grp_num= 2, rand_index= 1
  2759. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2760. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  2761. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2762. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  2763. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  2764. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2765. in math_function.cpp, line 730, tmp_value = 262144
  2766. in math_function.cpp line 656, gspt_info_1[i]= 6854148, gspt_info_2[i]= 0
  2767. definition of type is half.
  2768. gspt r[i] = 0.00195312 (Q_frac=27
  2769. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2770. In blob.cpp, blob.Q_frac= 27
  2771. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2772. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2773. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2774. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2775. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2776. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2777. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2778. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2779. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2780. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2781. in math_function.cpp, line 1227, tmp_value = 0
  2782. definition of type is half [FILL ZERO].
  2783. gspt r[i] = 0 (Q_frac=22
  2784. In blob.cpp, blob.Q_frac= 22
  2785. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2786. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  2787. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  2788. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2789. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  2790. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2791. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  2792. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  2793. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2794. in math_function.cpp, line 730, tmp_value = 1835008
  2795. in math_function.cpp line 656, gspt_info_1[i]= 1115270, gspt_info_2[i]= 0
  2796. definition of type is half.
  2797. gspt r[i] = 0.0136719 (Q_frac=27
  2798. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2799. In blob.cpp, blob.Q_frac= 27
  2800. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2801. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2802. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2803. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2804. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2805. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2806. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2807. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2808. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2809. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2810. in math_function.cpp, line 1227, tmp_value = 0
  2811. definition of type is half [FILL ZERO].
  2812. gspt r[i] = 0 (Q_frac=22
  2813. In blob.cpp, blob.Q_frac= 22
  2814. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2815. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  2816. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  2817. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  2818. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  2819. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2820. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2821. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  2822. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2823. in math_function.cpp, line 730, tmp_value = 0
  2824. in math_function.cpp line 656, gspt_info_1[i]= 1260883, gspt_info_2[i]= 0
  2825. definition of type is half.
  2826. gspt r[i] = 0 (Q_frac=28
  2827. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2828. In blob.cpp, blob.Q_frac= 27
  2829. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2830. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2831. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2832. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2833. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2834. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2835. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2836. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2837. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2838. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2839. in math_function.cpp, line 1227, tmp_value = 0
  2840. definition of type is half [FILL ZERO].
  2841. gspt r[i] = 0 (Q_frac=22
  2842. In blob.cpp, blob.Q_frac= 22
  2843. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2844. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  2845. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  2846. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  2847. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  2848. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2849. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  2850. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  2851. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2852. in math_function.cpp, line 730, tmp_value = -2228224
  2853. in math_function.cpp line 656, gspt_info_1[i]= 1583752, gspt_info_2[i]= 0
  2854. definition of type is half.
  2855. gspt r[i] = -0.00830078 (Q_frac=28
  2856. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2857. In blob.cpp, blob.Q_frac= 27
  2858. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2859. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2860. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2861. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2862. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2863. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2864. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2865. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2866. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2867. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2868. in math_function.cpp, line 1227, tmp_value = 0
  2869. definition of type is half [FILL ZERO].
  2870. gspt r[i] = 0 (Q_frac=22
  2871. In blob.cpp, blob.Q_frac= 22
  2872. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  2873. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  2874. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  2875. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2876. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  2877. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  2878. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2879. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  2880. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2881. in math_function.cpp, line 730, tmp_value = -1048576
  2882. in math_function.cpp line 656, gspt_info_1[i]= 13833745, gspt_info_2[i]= 0
  2883. definition of type is half.
  2884. gspt r[i] = -0.00390625 (Q_frac=28
  2885. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2886. In blob.cpp, blob.Q_frac= 27
  2887. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2888. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2889. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2890. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2891. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2892. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2893. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2894. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2895. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2896. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2897. in math_function.cpp, line 1227, tmp_value = 0
  2898. definition of type is half [FILL ZERO].
  2899. gspt r[i] = 0 (Q_frac=22
  2900. In blob.cpp, blob.Q_frac= 22
  2901. In blob.cpp, blob.Q_frac= 27
  2902. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  2903. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2904. in math_function.cpp line 565, grp_num= 2, rand_index= 1
  2905. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  2906. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  2907. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2908. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  2909. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  2910. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2911. in math_function.cpp, line 730, tmp_value = -917504
  2912. in math_function.cpp line 656, gspt_info_1[i]= 13932889, gspt_info_2[i]= 0
  2913. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2914. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2915. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2916. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2917. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2918. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2919. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2920. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2921. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2922. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2923. in math_function.cpp, line 1227, tmp_value = 0
  2924. definition of type is half [FILL ZERO].
  2925. gspt r[i] = 0 (Q_frac=22
  2926. In blob.cpp, blob.Q_frac= 22
  2927. definition of type is half.
  2928. gspt r[i] = -0.00341797 (Q_frac=28
  2929. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2930. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  2931. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  2932. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  2933. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  2934. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  2935. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2936. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  2937. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  2938. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2939. in math_function.cpp, line 730, tmp_value = 2228224
  2940. in math_function.cpp line 656, gspt_info_1[i]= 172894, gspt_info_2[i]= 0
  2941. definition of type is half.
  2942. gspt r[i] = 0.00830078 (Q_frac=28
  2943. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2944. In blob.cpp, blob.Q_frac= 27
  2945. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2946. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2947. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2948. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2949. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2950. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2951. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2952. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2953. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2954. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2955. in math_function.cpp, line 1227, tmp_value = 0
  2956. definition of type is half [FILL ZERO].
  2957. gspt r[i] = 0 (Q_frac=22
  2958. In blob.cpp, blob.Q_frac= 22
  2959. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  2960. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  2961. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  2962. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  2963. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  2964. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  2965. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  2966. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  2967. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2968. in math_function.cpp, line 730, tmp_value = -1966080
  2969. in math_function.cpp line 656, gspt_info_1[i]= 9777304, gspt_info_2[i]= 0
  2970. definition of type is half.
  2971. gspt r[i] = -0.00732422 (Q_frac=28
  2972. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  2973. In blob.cpp, blob.Q_frac= 28
  2974. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  2975. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  2976. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  2977. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  2978. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  2979. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  2980. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  2981. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  2982. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  2983. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  2984. in math_function.cpp, line 1227, tmp_value = 0
  2985. definition of type is half [FILL ZERO].
  2986. gspt r[i] = 0 (Q_frac=22
  2987. In blob.cpp, blob.Q_frac= 22
  2988. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  2989. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  2990. in math_function.cpp line 565, grp_num= 2, rand_index= 1
  2991. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  2992. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  2993. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  2994. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  2995. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  2996. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  2997. in math_function.cpp, line 730, tmp_value = 1048576
  2998. in math_function.cpp line 656, gspt_info_1[i]= 4751701, gspt_info_2[i]= 0
  2999. definition of type is half.
  3000. gspt r[i] = 0.00390625 (Q_frac=28
  3001. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3002. In blob.cpp, blob.Q_frac= 28
  3003. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3004. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3005. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3006. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3007. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3008. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3009. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3010. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3011. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3012. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3013. in math_function.cpp, line 1227, tmp_value = 0
  3014. definition of type is half [FILL ZERO].
  3015. gspt r[i] = 0 (Q_frac=22
  3016. In blob.cpp, blob.Q_frac= 22
  3017. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3018. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  3019. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3020. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3021. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  3022. in math_function.cpp line 565, grp_num= 5, rand_index= 6
  3023. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3024. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  3025. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3026. in math_function.cpp, line 730, tmp_value = 1179648
  3027. in math_function.cpp line 656, gspt_info_1[i]= 1125789, gspt_info_2[i]= 0
  3028. definition of type is half.
  3029. gspt r[i] = 0.00439453 (Q_frac=28
  3030. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3031. In blob.cpp, blob.Q_frac= 28
  3032. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3033. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3034. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3035. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3036. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3037. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3038. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3039. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3040. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3041. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3042. in math_function.cpp, line 1227, tmp_value = 0
  3043. definition of type is half [FILL ZERO].
  3044. gspt r[i] = 0 (Q_frac=22
  3045. In blob.cpp, blob.Q_frac= 22
  3046. In blob.cpp, blob.Q_frac= 28
  3047. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3048. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3049. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3050. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3051. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3052. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3053. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3054. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3055. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3056. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3057. in math_function.cpp, line 1227, tmp_value = 0
  3058. definition of type is half [FILL ZERO].
  3059. gspt r[i] = 0 (Q_frac=22
  3060. In blob.cpp, blob.Q_frac= 22
  3061. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  3062. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  3063. in math_function.cpp line 565, grp_num= 2, rand_index= 1
  3064. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3065. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  3066. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  3067. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3068. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  3069. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3070. in math_function.cpp, line 730, tmp_value = -917504
  3071. in math_function.cpp line 656, gspt_info_1[i]= 10788953, gspt_info_2[i]= 0
  3072. definition of type is half.
  3073. gspt r[i] = -0.00341797 (Q_frac=28
  3074. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3075. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  3076. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  3077. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3078. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3079. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  3080. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3081. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  3082. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  3083. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3084. in math_function.cpp, line 730, tmp_value = 524288
  3085. in math_function.cpp line 656, gspt_info_1[i]= 5057364, gspt_info_2[i]= 0
  3086. definition of type is half.
  3087. gspt r[i] = 0.00195312 (Q_frac=28
  3088. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3089. In blob.cpp, blob.Q_frac= 28
  3090. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3091. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3092. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3093. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3094. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3095. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3096. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3097. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3098. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3099. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3100. in math_function.cpp, line 1227, tmp_value = 0
  3101. definition of type is half [FILL ZERO].
  3102. gspt r[i] = 0 (Q_frac=22
  3103. In blob.cpp, blob.Q_frac= 22
  3104. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  3105. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3106. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3107. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  3108. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  3109. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  3110. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  3111. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  3112. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3113. in math_function.cpp, line 730, tmp_value = -1048576
  3114. in math_function.cpp line 656, gspt_info_1[i]= 11877137, gspt_info_2[i]= 0
  3115. definition of type is half.
  3116. gspt r[i] = -0.00390625 (Q_frac=28
  3117. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3118. In blob.cpp, blob.Q_frac= 28
  3119. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3120. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3121. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3122. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3123. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3124. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3125. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3126. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3127. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3128. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3129. in math_function.cpp, line 1227, tmp_value = 0
  3130. definition of type is half [FILL ZERO].
  3131. gspt r[i] = 0 (Q_frac=22
  3132. In blob.cpp, blob.Q_frac= 22
  3133. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3134. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3135. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3136. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  3137. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  3138. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3139. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  3140. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  3141. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3142. in math_function.cpp, line 730, tmp_value = -786432
  3143. in math_function.cpp line 656, gspt_info_1[i]= 1390914, gspt_info_2[i]= 0
  3144. definition of type is half.
  3145. gspt r[i] = -0.00292969 (Q_frac=28
  3146. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3147. In blob.cpp, blob.Q_frac= 28
  3148. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3149. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3150. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3151. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3152. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3153. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3154. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3155. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3156. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3157. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3158. in math_function.cpp, line 1227, tmp_value = 0
  3159. definition of type is half [FILL ZERO].
  3160. gspt r[i] = 0 (Q_frac=22
  3161. In blob.cpp, blob.Q_frac= 22
  3162. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  3163. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  3164. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  3165. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  3166. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  3167. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  3168. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  3169. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  3170. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3171. in math_function.cpp, line 730, tmp_value = -262144
  3172. in math_function.cpp line 656, gspt_info_1[i]= 5355523, gspt_info_2[i]= 0
  3173. definition of type is half.
  3174. gspt r[i] = -0.000976562 (Q_frac=28
  3175. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3176. In blob.cpp, blob.Q_frac= 28
  3177. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3178. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3179. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3180. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3181. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3182. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3183. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3184. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3185. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3186. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3187. in math_function.cpp, line 1227, tmp_value = 0
  3188. definition of type is half [FILL ZERO].
  3189. gspt r[i] = 0 (Q_frac=22
  3190. In blob.cpp, blob.Q_frac= 22
  3191. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3192. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  3193. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  3194. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  3195. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  3196. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3197. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3198. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  3199. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3200. in math_function.cpp, line 730, tmp_value = 131072
  3201. in math_function.cpp line 656, gspt_info_1[i]= 957275, gspt_info_2[i]= 0
  3202. definition of type is half.
  3203. gspt r[i] = 0.000976562 (Q_frac=27
  3204. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3205. In blob.cpp, blob.Q_frac= 27
  3206. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3207. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3208. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3209. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3210. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3211. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3212. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3213. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3214. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3215. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3216. in math_function.cpp, line 1227, tmp_value = 0
  3217. definition of type is half [FILL ZERO].
  3218. gspt r[i] = 0 (Q_frac=22
  3219. In blob.cpp, blob.Q_frac= 22
  3220. I0818 13:38:26.373353 18525 solver.cpp:174] Creating test net (#0) specified by net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  3221. W0818 13:38:26.373611 18525 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  3222. I0818 13:38:26.373821 18525 internal_thread.cpp:18] Starting 1 internal thread(s) on device 2
  3223. I0818 13:38:26.377493 18525 data_reader.cpp:59] Data Reader threads: 1, out queues: 1, depth: 13
  3224. I0818 13:38:26.377516 18525 internal_thread.cpp:18] Starting 1 internal thread(s) on device 2
  3225. I0818 13:38:26.379009 18741 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_test_lmdb
  3226. I0818 13:38:26.380197 18525 data_layer.cpp:200] (n0.d2.r2) Output data size: 13, 3, 32, 32
  3227. I0818 13:38:26.381037 18525 internal_thread.cpp:18] Starting 1 internal thread(s) on device 2
  3228. I0818 13:38:26.382710 18742 data_layer.cpp:105] (n0.d2.r2) Parser threads: 1
  3229. I0818 13:38:26.382730 18742 data_layer.cpp:107] (n0.d2.r2) Transformer threads: 1
  3230. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3231. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  3232. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  3233. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  3234. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  3235. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  3236. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  3237. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  3238. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3239. in math_function.cpp, line 730, tmp_value = -2097152
  3240. in math_function.cpp line 656, gspt_info_1[i]= 641040, gspt_info_2[i]= 0
  3241. definition of type is half.
  3242. gspt r[i] = -0.125 (Q_frac=24
  3243. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3244. In blob.cpp, blob.Q_frac= 24
  3245. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3246. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3247. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3248. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3249. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3250. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3251. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3252. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3253. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3254. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3255. in math_function.cpp, line 1227, tmp_value = 0
  3256. definition of type is half [FILL ZERO].
  3257. gspt r[i] = 0 (Q_frac=22
  3258. In blob.cpp, blob.Q_frac= 22
  3259. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  3260. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  3261. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  3262. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  3263. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  3264. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3265. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  3266. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  3267. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3268. in math_function.cpp, line 730, tmp_value = -524288
  3269. in math_function.cpp line 656, gspt_info_1[i]= 4834130, gspt_info_2[i]= 0
  3270. definition of type is half.
  3271. gspt r[i] = -0.00390625 (Q_frac=27
  3272. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3273. In blob.cpp, blob.Q_frac= 27
  3274. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3275. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3276. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3277. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3278. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3279. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3280. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3281. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3282. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3283. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3284. in math_function.cpp, line 1227, tmp_value = 0
  3285. definition of type is half [FILL ZERO].
  3286. gspt r[i] = 0 (Q_frac=22
  3287. In blob.cpp, blob.Q_frac= 22
  3288. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  3289. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  3290. in math_function.cpp line 565, grp_num= 2, rand_index= 1
  3291. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  3292. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  3293. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  3294. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  3295. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  3296. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3297. in math_function.cpp, line 730, tmp_value = 786432
  3298. in math_function.cpp line 656, gspt_info_1[i]= 12882084, gspt_info_2[i]= 0
  3299. definition of type is half.
  3300. gspt r[i] = 0.00585938 (Q_frac=27
  3301. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3302. In blob.cpp, blob.Q_frac= 27
  3303. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3304. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3305. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3306. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3307. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3308. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3309. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3310. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3311. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3312. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3313. in math_function.cpp, line 1227, tmp_value = 0
  3314. definition of type is half [FILL ZERO].
  3315. gspt r[i] = 0 (Q_frac=22
  3316. In blob.cpp, blob.Q_frac= 22
  3317. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3318. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3319. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  3320. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  3321. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  3322. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3323. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3324. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  3325. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3326. in math_function.cpp, line 730, tmp_value = -1966080
  3327. in math_function.cpp line 656, gspt_info_1[i]= 1442136, gspt_info_2[i]= 0
  3328. definition of type is half.
  3329. gspt r[i] = -0.0146484 (Q_frac=27
  3330. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3331. In blob.cpp, blob.Q_frac= 27
  3332. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3333. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3334. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3335. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3336. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3337. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3338. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3339. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3340. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3341. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3342. in math_function.cpp, line 1227, tmp_value = 0
  3343. definition of type is half [FILL ZERO].
  3344. gspt r[i] = 0 (Q_frac=22
  3345. In blob.cpp, blob.Q_frac= 22
  3346. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  3347. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  3348. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  3349. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  3350. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  3351. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  3352. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  3353. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  3354. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3355. in math_function.cpp, line 730, tmp_value = 262144
  3356. in math_function.cpp line 656, gspt_info_1[i]= 13272580, gspt_info_2[i]= 0
  3357. definition of type is half.
  3358. gspt r[i] = 0.00195312 (Q_frac=27
  3359. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3360. In blob.cpp, blob.Q_frac= 28
  3361. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3362. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3363. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3364. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3365. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3366. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3367. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3368. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3369. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3370. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3371. in math_function.cpp, line 1227, tmp_value = 0
  3372. definition of type is half [FILL ZERO].
  3373. gspt r[i] = 0 (Q_frac=22
  3374. In blob.cpp, blob.Q_frac= 22
  3375. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  3376. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3377. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  3378. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3379. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  3380. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  3381. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  3382. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  3383. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3384. in math_function.cpp, line 730, tmp_value = -262144
  3385. in math_function.cpp line 656, gspt_info_1[i]= 9709218, gspt_info_2[i]= 0
  3386. definition of type is half.
  3387. gspt r[i] = -0.00195312 (Q_frac=27
  3388. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3389. In blob.cpp, blob.Q_frac= 27
  3390. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3391. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3392. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3393. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3394. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3395. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3396. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3397. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3398. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3399. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3400. in math_function.cpp, line 1227, tmp_value = 0
  3401. definition of type is half [FILL ZERO].
  3402. gspt r[i] = 0 (Q_frac=22
  3403. In blob.cpp, blob.Q_frac= 22
  3404. I0818 13:38:45.458088 18528 solver.cpp:174] Creating test net (#0) specified by net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  3405. W0818 13:38:45.458241 18528 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  3406. I0818 13:38:45.458479 18528 internal_thread.cpp:18] Starting 1 internal thread(s) on device 5
  3407. I0818 13:38:45.461125 18528 data_reader.cpp:59] Data Reader threads: 1, out queues: 1, depth: 13
  3408. I0818 13:38:45.461150 18528 internal_thread.cpp:18] Starting 1 internal thread(s) on device 5
  3409. I0818 13:38:45.462862 18775 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_test_lmdb
  3410. I0818 13:38:45.464134 18528 data_layer.cpp:200] (n0.d5.r5) Output data size: 13, 3, 32, 32
  3411. I0818 13:38:45.465052 18528 internal_thread.cpp:18] Starting 1 internal thread(s) on device 5
  3412. I0818 13:38:45.466584 18776 data_layer.cpp:105] (n0.d5.r5) Parser threads: 1
  3413. I0818 13:38:45.466599 18776 data_layer.cpp:107] (n0.d5.r5) Transformer threads: 1
  3414. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3415. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  3416. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  3417. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  3418. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  3419. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  3420. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3421. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  3422. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3423. in math_function.cpp, line 730, tmp_value = 1179648
  3424. in math_function.cpp line 656, gspt_info_1[i]= 214109, gspt_info_2[i]= 0
  3425. definition of type is half.
  3426. gspt r[i] = 0.0703125 (Q_frac=24
  3427. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3428. In blob.cpp, blob.Q_frac= 24
  3429. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3430. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3431. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3432. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3433. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3434. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3435. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3436. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3437. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3438. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3439. in math_function.cpp, line 1227, tmp_value = 0
  3440. definition of type is half [FILL ZERO].
  3441. gspt r[i] = 0 (Q_frac=22
  3442. In blob.cpp, blob.Q_frac= 22
  3443. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  3444. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  3445. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3446. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  3447. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  3448. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  3449. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  3450. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  3451. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3452. in math_function.cpp, line 730, tmp_value = 2097152
  3453. in math_function.cpp line 656, gspt_info_1[i]= 8985174, gspt_info_2[i]= 0
  3454. definition of type is half.
  3455. gspt r[i] = 0.015625 (Q_frac=27
  3456. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3457. In blob.cpp, blob.Q_frac= 27
  3458. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3459. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3460. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3461. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3462. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3463. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3464. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3465. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3466. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3467. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3468. in math_function.cpp, line 1227, tmp_value = 0
  3469. definition of type is half [FILL ZERO].
  3470. gspt r[i] = 0 (Q_frac=22
  3471. In blob.cpp, blob.Q_frac= 22
  3472. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  3473. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  3474. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  3475. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  3476. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  3477. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3478. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3479. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  3480. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3481. in math_function.cpp, line 730, tmp_value = 1179648
  3482. in math_function.cpp line 656, gspt_info_1[i]= 12751709, gspt_info_2[i]= 0
  3483. definition of type is half.
  3484. gspt r[i] = 0.00878906 (Q_frac=27
  3485. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3486. In blob.cpp, blob.Q_frac= 28
  3487. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3488. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3489. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3490. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3491. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3492. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3493. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3494. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3495. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3496. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3497. in math_function.cpp, line 1227, tmp_value = 0
  3498. definition of type is half [FILL ZERO].
  3499. gspt r[i] = 0 (Q_frac=22
  3500. In blob.cpp, blob.Q_frac= 22
  3501. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  3502. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3503. in math_function.cpp line 565, grp_num= 2, rand_index= 1
  3504. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  3505. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  3506. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3507. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3508. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  3509. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3510. in math_function.cpp, line 730, tmp_value = -917504
  3511. in math_function.cpp line 656, gspt_info_1[i]= 9737049, gspt_info_2[i]= 0
  3512. definition of type is half.
  3513. gspt r[i] = -0.00683594 (Q_frac=27
  3514. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3515. In blob.cpp, blob.Q_frac= 27
  3516. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3517. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3518. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3519. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3520. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3521. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3522. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3523. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3524. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3525. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3526. in math_function.cpp, line 1227, tmp_value = 0
  3527. definition of type is half [FILL ZERO].
  3528. gspt r[i] = 0 (Q_frac=22
  3529. In blob.cpp, blob.Q_frac= 22
  3530. I0818 13:38:52.400616 18527 solver.cpp:174] Creating test net (#0) specified by net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  3531. W0818 13:38:52.400733 18527 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  3532. I0818 13:38:52.400933 18527 internal_thread.cpp:18] Starting 1 internal thread(s) on device 4
  3533. I0818 13:38:52.402825 18527 data_reader.cpp:59] Data Reader threads: 1, out queues: 1, depth: 13
  3534. I0818 13:38:52.402848 18527 internal_thread.cpp:18] Starting 1 internal thread(s) on device 4
  3535. I0818 13:38:52.404686 18797 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_test_lmdb
  3536. I0818 13:38:52.405822 18527 data_layer.cpp:200] (n0.d4.r4) Output data size: 13, 3, 32, 32
  3537. I0818 13:38:52.406626 18527 internal_thread.cpp:18] Starting 1 internal thread(s) on device 4
  3538. I0818 13:38:52.408490 18798 data_layer.cpp:105] (n0.d4.r4) Parser threads: 1
  3539. I0818 13:38:52.408530 18798 data_layer.cpp:107] (n0.d4.r4) Transformer threads: 1
  3540. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  3541. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  3542. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3543. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  3544. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  3545. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  3546. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3547. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  3548. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3549. in math_function.cpp, line 730, tmp_value = -917504
  3550. in math_function.cpp line 656, gspt_info_1[i]= 11076889, gspt_info_2[i]= 0
  3551. definition of type is half.
  3552. gspt r[i] = -0.0546875 (Q_frac=24
  3553. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3554. In blob.cpp, blob.Q_frac= 24
  3555. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3556. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3557. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3558. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3559. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3560. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3561. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3562. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3563. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3564. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3565. in math_function.cpp, line 1227, tmp_value = 0
  3566. definition of type is half [FILL ZERO].
  3567. gspt r[i] = 0 (Q_frac=22
  3568. In blob.cpp, blob.Q_frac= 22
  3569. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  3570. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  3571. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3572. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  3573. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  3574. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  3575. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3576. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  3577. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3578. in math_function.cpp, line 730, tmp_value = -1966080
  3579. in math_function.cpp line 656, gspt_info_1[i]= 5063832, gspt_info_2[i]= 0
  3580. definition of type is half.
  3581. gspt r[i] = -0.0146484 (Q_frac=27
  3582. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3583. In blob.cpp, blob.Q_frac= 27
  3584. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3585. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3586. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3587. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3588. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3589. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3590. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3591. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3592. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3593. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3594. in math_function.cpp, line 1227, tmp_value = 0
  3595. definition of type is half [FILL ZERO].
  3596. gspt r[i] = 0 (Q_frac=22
  3597. In blob.cpp, blob.Q_frac= 22
  3598. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  3599. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  3600. in math_function.cpp line 565, grp_num= 2, rand_index= 1
  3601. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  3602. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  3603. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3604. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  3605. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  3606. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3607. in math_function.cpp, line 730, tmp_value = 1048576
  3608. in math_function.cpp line 656, gspt_info_1[i]= 4764501, gspt_info_2[i]= 0
  3609. definition of type is half.
  3610. gspt r[i] = 0.0078125 (Q_frac=27
  3611. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3612. In blob.cpp, blob.Q_frac= 27
  3613. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3614. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3615. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3616. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3617. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3618. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3619. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3620. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3621. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3622. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3623. in math_function.cpp, line 1227, tmp_value = 0
  3624. definition of type is half [FILL ZERO].
  3625. gspt r[i] = 0 (Q_frac=22
  3626. In blob.cpp, blob.Q_frac= 22
  3627. in math_function.cpp line 565, grp_num= 0, rand_index= 3
  3628. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3629. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  3630. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  3631. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  3632. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3633. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  3634. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  3635. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3636. in math_function.cpp, line 730, tmp_value = 0
  3637. in math_function.cpp line 656, gspt_info_1[i]= 7609683, gspt_info_2[i]= 0
  3638. definition of type is half.
  3639. gspt r[i] = 0 (Q_frac=27
  3640. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3641. In blob.cpp, blob.Q_frac= 28
  3642. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3643. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3644. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3645. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3646. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3647. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3648. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3649. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3650. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3651. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3652. in math_function.cpp, line 1227, tmp_value = 0
  3653. definition of type is half [FILL ZERO].
  3654. gspt r[i] = 0 (Q_frac=22
  3655. In blob.cpp, blob.Q_frac= 22
  3656. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  3657. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  3658. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  3659. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  3660. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  3661. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  3662. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  3663. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  3664. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3665. in math_function.cpp, line 730, tmp_value = 393216
  3666. in math_function.cpp line 656, gspt_info_1[i]= 4208844, gspt_info_2[i]= 0
  3667. definition of type is half.
  3668. gspt r[i] = 0.00292969 (Q_frac=27
  3669. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3670. In blob.cpp, blob.Q_frac= 27
  3671. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3672. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3673. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3674. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3675. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3676. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3677. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3678. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3679. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3680. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3681. in math_function.cpp, line 1227, tmp_value = 0
  3682. definition of type is half [FILL ZERO].
  3683. gspt r[i] = 0 (Q_frac=22
  3684. In blob.cpp, blob.Q_frac= 22
  3685. I0818 13:38:55.580049 18529 solver.cpp:174] Creating test net (#0) specified by net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  3686. W0818 13:38:55.580170 18529 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  3687. I0818 13:38:55.580372 18529 internal_thread.cpp:18] Starting 1 internal thread(s) on device 6
  3688. I0818 13:38:55.582173 18529 data_reader.cpp:59] Data Reader threads: 1, out queues: 1, depth: 13
  3689. I0818 13:38:55.582193 18529 internal_thread.cpp:18] Starting 1 internal thread(s) on device 6
  3690. I0818 13:38:55.583535 18809 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_test_lmdb
  3691. I0818 13:38:55.584699 18529 data_layer.cpp:200] (n0.d6.r6) Output data size: 13, 3, 32, 32
  3692. I0818 13:38:55.585515 18529 internal_thread.cpp:18] Starting 1 internal thread(s) on device 6
  3693. I0818 13:38:55.587839 18810 data_layer.cpp:105] (n0.d6.r6) Parser threads: 1
  3694. I0818 13:38:55.587865 18810 data_layer.cpp:107] (n0.d6.r6) Transformer threads: 1
  3695. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  3696. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  3697. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  3698. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3699. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  3700. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  3701. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  3702. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  3703. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3704. in math_function.cpp, line 730, tmp_value = -2359296
  3705. in math_function.cpp line 656, gspt_info_1[i]= 13248128, gspt_info_2[i]= 0
  3706. definition of type is half.
  3707. gspt r[i] = -0.140625 (Q_frac=24
  3708. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3709. In blob.cpp, blob.Q_frac= 24
  3710. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3711. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3712. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3713. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3714. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3715. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3716. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3717. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3718. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3719. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3720. in math_function.cpp, line 1227, tmp_value = 0
  3721. definition of type is half [FILL ZERO].
  3722. gspt r[i] = 0 (Q_frac=22
  3723. In blob.cpp, blob.Q_frac= 22
  3724. in math_function.cpp line 565, grp_num= 0, rand_index= 1
  3725. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  3726. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  3727. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3728. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  3729. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3730. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  3731. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  3732. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3733. in math_function.cpp, line 730, tmp_value = 262144
  3734. in math_function.cpp line 656, gspt_info_1[i]= 3351364, gspt_info_2[i]= 0
  3735. definition of type is half.
  3736. gspt r[i] = 0.00195312 (Q_frac=27
  3737. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3738. In blob.cpp, blob.Q_frac= 28
  3739. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3740. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3741. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3742. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3743. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3744. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3745. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3746. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3747. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3748. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3749. in math_function.cpp, line 1227, tmp_value = 0
  3750. definition of type is half [FILL ZERO].
  3751. gspt r[i] = 0 (Q_frac=22
  3752. In blob.cpp, blob.Q_frac= 22
  3753. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  3754. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  3755. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  3756. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  3757. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  3758. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3759. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  3760. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  3761. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3762. in math_function.cpp, line 730, tmp_value = -1310720
  3763. in math_function.cpp line 656, gspt_info_1[i]= 12584769, gspt_info_2[i]= 0
  3764. definition of type is half.
  3765. gspt r[i] = -0.00976562 (Q_frac=27
  3766. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3767. In blob.cpp, blob.Q_frac= 27
  3768. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3769. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3770. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3771. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3772. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3773. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3774. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3775. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3776. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3777. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3778. in math_function.cpp, line 1227, tmp_value = 0
  3779. definition of type is half [FILL ZERO].
  3780. gspt r[i] = 0 (Q_frac=22
  3781. In blob.cpp, blob.Q_frac= 22
  3782. I0818 13:38:56.121636 18526 solver.cpp:174] Creating test net (#0) specified by net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  3783. W0818 13:38:56.121800 18526 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  3784. I0818 13:38:56.122026 18526 internal_thread.cpp:18] Starting 1 internal thread(s) on device 3
  3785. I0818 13:38:56.124115 18526 data_reader.cpp:59] Data Reader threads: 1, out queues: 1, depth: 13
  3786. I0818 13:38:56.124152 18526 internal_thread.cpp:18] Starting 1 internal thread(s) on device 3
  3787. I0818 13:38:56.125921 18818 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_test_lmdb
  3788. I0818 13:38:56.127220 18526 data_layer.cpp:200] (n0.d3.r3) Output data size: 13, 3, 32, 32
  3789. I0818 13:38:56.128099 18526 internal_thread.cpp:18] Starting 1 internal thread(s) on device 3
  3790. I0818 13:38:56.129609 18819 data_layer.cpp:105] (n0.d3.r3) Parser threads: 1
  3791. I0818 13:38:56.129626 18819 data_layer.cpp:107] (n0.d3.r3) Transformer threads: 1
  3792. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  3793. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3794. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3795. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3796. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  3797. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  3798. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  3799. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  3800. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3801. in math_function.cpp, line 730, tmp_value = 786432
  3802. in math_function.cpp line 656, gspt_info_1[i]= 11870884, gspt_info_2[i]= 0
  3803. definition of type is half.
  3804. gspt r[i] = 0.046875 (Q_frac=24
  3805. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3806. In blob.cpp, blob.Q_frac= 24
  3807. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3808. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3809. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3810. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3811. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3812. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3813. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3814. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3815. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3816. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3817. in math_function.cpp, line 1227, tmp_value = 0
  3818. definition of type is half [FILL ZERO].
  3819. gspt r[i] = 0 (Q_frac=22
  3820. In blob.cpp, blob.Q_frac= 22
  3821. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3822. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3823. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3824. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3825. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  3826. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  3827. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  3828. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  3829. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3830. in math_function.cpp, line 730, tmp_value = -2228224
  3831. in math_function.cpp line 656, gspt_info_1[i]= 1386504, gspt_info_2[i]= 0
  3832. definition of type is half.
  3833. gspt r[i] = -0.0166016 (Q_frac=27
  3834. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3835. In blob.cpp, blob.Q_frac= 28
  3836. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3837. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3838. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3839. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3840. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3841. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3842. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3843. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3844. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3845. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3846. in math_function.cpp, line 1227, tmp_value = 0
  3847. definition of type is half [FILL ZERO].
  3848. gspt r[i] = 0 (Q_frac=22
  3849. In blob.cpp, blob.Q_frac= 22
  3850. in math_function.cpp line 565, grp_num= 0, rand_index= 3
  3851. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3852. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  3853. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  3854. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  3855. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  3856. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3857. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  3858. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3859. in math_function.cpp, line 730, tmp_value = -1966080
  3860. in math_function.cpp line 656, gspt_info_1[i]= 7675928, gspt_info_2[i]= 0
  3861. definition of type is half.
  3862. gspt r[i] = -0.0146484 (Q_frac=27
  3863. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3864. In blob.cpp, blob.Q_frac= 27
  3865. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3866. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3867. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3868. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3869. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3870. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3871. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3872. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3873. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3874. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3875. in math_function.cpp, line 1227, tmp_value = 0
  3876. definition of type is half [FILL ZERO].
  3877. gspt r[i] = 0 (Q_frac=22
  3878. In blob.cpp, blob.Q_frac= 22
  3879. I0818 13:38:57.086112 18524 solver.cpp:174] Creating test net (#0) specified by net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  3880. W0818 13:38:57.086496 18524 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  3881. I0818 13:38:57.086737 18524 internal_thread.cpp:18] Starting 1 internal thread(s) on device 1
  3882. I0818 13:38:57.089479 18524 data_reader.cpp:59] Data Reader threads: 1, out queues: 1, depth: 13
  3883. I0818 13:38:57.089515 18524 internal_thread.cpp:18] Starting 1 internal thread(s) on device 1
  3884. I0818 13:38:57.091111 18821 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_test_lmdb
  3885. I0818 13:38:57.092270 18524 data_layer.cpp:200] (n0.d1.r1) Output data size: 13, 3, 32, 32
  3886. I0818 13:38:57.093107 18524 internal_thread.cpp:18] Starting 1 internal thread(s) on device 1
  3887. I0818 13:38:57.094491 18822 data_layer.cpp:105] (n0.d1.r1) Parser threads: 1
  3888. I0818 13:38:57.094508 18822 data_layer.cpp:107] (n0.d1.r1) Transformer threads: 1
  3889. in math_function.cpp line 565, grp_num= 0, rand_index= 1
  3890. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  3891. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  3892. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  3893. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  3894. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  3895. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3896. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  3897. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3898. in math_function.cpp, line 730, tmp_value = 131072
  3899. in math_function.cpp line 656, gspt_info_1[i]= 3674395, gspt_info_2[i]= 0
  3900. definition of type is half.
  3901. gspt r[i] = 0.0078125 (Q_frac=24
  3902. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3903. In blob.cpp, blob.Q_frac= 24
  3904. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3905. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3906. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3907. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3908. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3909. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3910. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3911. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3912. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3913. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3914. in math_function.cpp, line 1227, tmp_value = 0
  3915. definition of type is half [FILL ZERO].
  3916. gspt r[i] = 0 (Q_frac=22
  3917. In blob.cpp, blob.Q_frac= 22
  3918. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  3919. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3920. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  3921. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  3922. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  3923. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  3924. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  3925. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  3926. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3927. in math_function.cpp, line 730, tmp_value = -917504
  3928. in math_function.cpp line 656, gspt_info_1[i]= 9811289, gspt_info_2[i]= 0
  3929. definition of type is half.
  3930. gspt r[i] = -0.00683594 (Q_frac=27
  3931. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3932. In blob.cpp, blob.Q_frac= 28
  3933. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3934. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3935. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3936. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3937. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3938. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3939. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3940. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3941. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3942. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3943. in math_function.cpp, line 1227, tmp_value = 0
  3944. definition of type is half [FILL ZERO].
  3945. gspt r[i] = 0 (Q_frac=22
  3946. In blob.cpp, blob.Q_frac= 22
  3947. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3948. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  3949. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  3950. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  3951. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  3952. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  3953. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  3954. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  3955. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3956. in math_function.cpp, line 730, tmp_value = -2228224
  3957. in math_function.cpp line 656, gspt_info_1[i]= 1514120, gspt_info_2[i]= 0
  3958. definition of type is half.
  3959. gspt r[i] = -0.0166016 (Q_frac=27
  3960. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  3961. In blob.cpp, blob.Q_frac= 27
  3962. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  3963. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  3964. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  3965. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  3966. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  3967. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  3968. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  3969. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  3970. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  3971. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  3972. in math_function.cpp, line 1227, tmp_value = 0
  3973. definition of type is half [FILL ZERO].
  3974. gspt r[i] = 0 (Q_frac=22
  3975. In blob.cpp, blob.Q_frac= 22
  3976. I0818 13:38:57.983455 18530 solver.cpp:174] Creating test net (#0) specified by net file: examples/cifar10/cifar10_vgg7_BN.prototxt
  3977. W0818 13:38:57.983579 18530 parallel.cpp:319] Batch size must be divisible by the number of solvers (GPUs): it's been adjusted from 100 to 104
  3978. I0818 13:38:57.983788 18530 internal_thread.cpp:18] Starting 1 internal thread(s) on device 7
  3979. I0818 13:38:57.985976 18530 data_reader.cpp:59] Data Reader threads: 1, out queues: 1, depth: 13
  3980. I0818 13:38:57.985997 18530 internal_thread.cpp:18] Starting 1 internal thread(s) on device 7
  3981. I0818 13:38:57.987715 18824 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_test_lmdb
  3982. I0818 13:38:57.988878 18530 data_layer.cpp:200] (n0.d7.r7) Output data size: 13, 3, 32, 32
  3983. I0818 13:38:57.989739 18530 internal_thread.cpp:18] Starting 1 internal thread(s) on device 7
  3984. I0818 13:38:57.991744 18825 data_layer.cpp:105] (n0.d7.r7) Parser threads: 1
  3985. I0818 13:38:57.991770 18825 data_layer.cpp:107] (n0.d7.r7) Transformer threads: 1
  3986. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  3987. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  3988. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  3989. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  3990. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  3991. in math_function.cpp line 565, grp_num= 5, rand_index= 6
  3992. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  3993. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  3994. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  3995. in math_function.cpp, line 730, tmp_value = -1048576
  3996. in math_function.cpp line 656, gspt_info_1[i]= 1722769, gspt_info_2[i]= 0
  3997. definition of type is half.
  3998. gspt r[i] = -0.0625 (Q_frac=24
  3999. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4000. In blob.cpp, blob.Q_frac= 24
  4001. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4002. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4003. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4004. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4005. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4006. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4007. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4008. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4009. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4010. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4011. in math_function.cpp, line 1227, tmp_value = 0
  4012. definition of type is half [FILL ZERO].
  4013. gspt r[i] = 0 (Q_frac=22
  4014. In blob.cpp, blob.Q_frac= 22
  4015. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  4016. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  4017. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  4018. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4019. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  4020. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  4021. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  4022. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  4023. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4024. in math_function.cpp, line 730, tmp_value = -1048576
  4025. in math_function.cpp line 656, gspt_info_1[i]= 8401553, gspt_info_2[i]= 0
  4026. definition of type is half.
  4027. gspt r[i] = -0.0078125 (Q_frac=27
  4028. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4029. In blob.cpp, blob.Q_frac= 27
  4030. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4031. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4032. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4033. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4034. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4035. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4036. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4037. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4038. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4039. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4040. in math_function.cpp, line 1227, tmp_value = 0
  4041. definition of type is half [FILL ZERO].
  4042. gspt r[i] = 0 (Q_frac=22
  4043. In blob.cpp, blob.Q_frac= 22
  4044. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  4045. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  4046. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  4047. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  4048. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  4049. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  4050. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4051. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  4052. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4053. in math_function.cpp, line 730, tmp_value = -917504
  4054. in math_function.cpp line 656, gspt_info_1[i]= 10882393, gspt_info_2[i]= 0
  4055. definition of type is half.
  4056. gspt r[i] = -0.00683594 (Q_frac=27
  4057. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4058. In blob.cpp, blob.Q_frac= 27
  4059. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4060. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4061. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4062. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4063. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4064. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4065. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4066. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4067. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4068. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4069. in math_function.cpp, line 1227, tmp_value = 0
  4070. definition of type is half [FILL ZERO].
  4071. gspt r[i] = 0 (Q_frac=22
  4072. In blob.cpp, blob.Q_frac= 22
  4073. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  4074. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  4075. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  4076. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4077. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  4078. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  4079. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  4080. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  4081. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4082. in math_function.cpp, line 730, tmp_value = 786432
  4083. in math_function.cpp line 656, gspt_info_1[i]= 12949540, gspt_info_2[i]= 0
  4084. definition of type is half.
  4085. gspt r[i] = 0.00585938 (Q_frac=27
  4086. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4087. In blob.cpp, blob.Q_frac= 27
  4088. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4089. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4090. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4091. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4092. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4093. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4094. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4095. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4096. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4097. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4098. in math_function.cpp, line 1227, tmp_value = 0
  4099. definition of type is half [FILL ZERO].
  4100. gspt r[i] = 0 (Q_frac=22
  4101. In blob.cpp, blob.Q_frac= 22
  4102. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4103. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  4104. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  4105. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4106. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  4107. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  4108. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4109. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  4110. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4111. in math_function.cpp, line 730, tmp_value = 917504
  4112. in math_function.cpp line 656, gspt_info_1[i]= 269645, gspt_info_2[i]= 0
  4113. definition of type is half.
  4114. gspt r[i] = 0.00683594 (Q_frac=27
  4115. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4116. In blob.cpp, blob.Q_frac= 27
  4117. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4118. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4119. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4120. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4121. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4122. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4123. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4124. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4125. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4126. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4127. in math_function.cpp, line 1227, tmp_value = 0
  4128. definition of type is half [FILL ZERO].
  4129. gspt r[i] = 0 (Q_frac=22
  4130. In blob.cpp, blob.Q_frac= 22
  4131. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4132. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  4133. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  4134. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4135. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  4136. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  4137. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4138. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  4139. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4140. in math_function.cpp, line 730, tmp_value = 2228224
  4141. in math_function.cpp line 656, gspt_info_1[i]= 1675870, gspt_info_2[i]= 0
  4142. definition of type is half.
  4143. gspt r[i] = 0.0166016 (Q_frac=27
  4144. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4145. In blob.cpp, blob.Q_frac= 27
  4146. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4147. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4148. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4149. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4150. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4151. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4152. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4153. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4154. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4155. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4156. in math_function.cpp, line 1227, tmp_value = 0
  4157. definition of type is half [FILL ZERO].
  4158. gspt r[i] = 0 (Q_frac=22
  4159. In blob.cpp, blob.Q_frac= 22
  4160. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  4161. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  4162. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  4163. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4164. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  4165. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  4166. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  4167. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  4168. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4169. in math_function.cpp, line 730, tmp_value = -1310720
  4170. in math_function.cpp line 656, gspt_info_1[i]= 4594369, gspt_info_2[i]= 0
  4171. definition of type is half.
  4172. gspt r[i] = -0.00976562 (Q_frac=27
  4173. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4174. In blob.cpp, blob.Q_frac= 27
  4175. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4176. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4177. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4178. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4179. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4180. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4181. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4182. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4183. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4184. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4185. in math_function.cpp, line 1227, tmp_value = 0
  4186. definition of type is half [FILL ZERO].
  4187. gspt r[i] = 0 (Q_frac=22
  4188. In blob.cpp, blob.Q_frac= 22
  4189. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  4190. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  4191. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  4192. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  4193. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  4194. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  4195. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4196. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  4197. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4198. in math_function.cpp, line 730, tmp_value = -2228224
  4199. in math_function.cpp line 656, gspt_info_1[i]= 14237960, gspt_info_2[i]= 0
  4200. definition of type is half.
  4201. gspt r[i] = -0.0166016 (Q_frac=27
  4202. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4203. In blob.cpp, blob.Q_frac= 27
  4204. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4205. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4206. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4207. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4208. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4209. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4210. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4211. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4212. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4213. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4214. in math_function.cpp, line 1227, tmp_value = 0
  4215. definition of type is half [FILL ZERO].
  4216. gspt r[i] = 0 (Q_frac=22
  4217. In blob.cpp, blob.Q_frac= 22
  4218. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4219. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  4220. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  4221. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4222. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  4223. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  4224. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4225. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  4226. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4227. in math_function.cpp, line 730, tmp_value = 1179648
  4228. in math_function.cpp line 656, gspt_info_1[i]= 432925, gspt_info_2[i]= 0
  4229. definition of type is half.
  4230. gspt r[i] = 0.00878906 (Q_frac=27
  4231. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4232. In blob.cpp, blob.Q_frac= 27
  4233. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4234. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4235. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4236. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4237. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4238. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4239. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4240. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4241. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4242. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4243. in math_function.cpp, line 1227, tmp_value = 0
  4244. definition of type is half [FILL ZERO].
  4245. gspt r[i] = 0 (Q_frac=22
  4246. In blob.cpp, blob.Q_frac= 22
  4247. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4248. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  4249. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  4250. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4251. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  4252. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  4253. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  4254. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  4255. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4256. in math_function.cpp, line 730, tmp_value = -262144
  4257. in math_function.cpp line 656, gspt_info_1[i]= 595203, gspt_info_2[i]= 0
  4258. definition of type is half.
  4259. gspt r[i] = -0.00195312 (Q_frac=27
  4260. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4261. In blob.cpp, blob.Q_frac= 27
  4262. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4263. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4264. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4265. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4266. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4267. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4268. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4269. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4270. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4271. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4272. in math_function.cpp, line 1227, tmp_value = 0
  4273. definition of type is half [FILL ZERO].
  4274. gspt r[i] = 0 (Q_frac=22
  4275. In blob.cpp, blob.Q_frac= 22
  4276. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  4277. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  4278. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  4279. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4280. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  4281. in math_function.cpp line 565, grp_num= 5, rand_index= 6
  4282. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  4283. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  4284. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4285. in math_function.cpp, line 730, tmp_value = -524288
  4286. in math_function.cpp line 656, gspt_info_1[i]= 4306322, gspt_info_2[i]= 0
  4287. definition of type is half.
  4288. gspt r[i] = -0.00195312 (Q_frac=28
  4289. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4290. In blob.cpp, blob.Q_frac= 27
  4291. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4292. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4293. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4294. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4295. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4296. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4297. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4298. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4299. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4300. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4301. in math_function.cpp, line 1227, tmp_value = 0
  4302. definition of type is half [FILL ZERO].
  4303. gspt r[i] = 0 (Q_frac=22
  4304. In blob.cpp, blob.Q_frac= 22
  4305. in math_function.cpp line 565, grp_num= 0, rand_index= 1
  4306. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  4307. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  4308. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4309. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  4310. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  4311. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4312. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  4313. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4314. in math_function.cpp, line 730, tmp_value = -393216
  4315. in math_function.cpp line 656, gspt_info_1[i]= 2569562, gspt_info_2[i]= 0
  4316. definition of type is half.
  4317. gspt r[i] = -0.00292969 (Q_frac=27
  4318. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4319. In blob.cpp, blob.Q_frac= 27
  4320. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4321. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4322. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4323. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4324. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4325. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4326. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4327. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4328. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4329. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4330. in math_function.cpp, line 1227, tmp_value = 0
  4331. definition of type is half [FILL ZERO].
  4332. gspt r[i] = 0 (Q_frac=22
  4333. In blob.cpp, blob.Q_frac= 22
  4334. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4335. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  4336. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  4337. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4338. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  4339. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  4340. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4341. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  4342. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4343. in math_function.cpp, line 730, tmp_value = -655360
  4344. in math_function.cpp line 656, gspt_info_1[i]= 1389322, gspt_info_2[i]= 0
  4345. definition of type is half.
  4346. gspt r[i] = -0.00488281 (Q_frac=27
  4347. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4348. In blob.cpp, blob.Q_frac= 27
  4349. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4350. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4351. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4352. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4353. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4354. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4355. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4356. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4357. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4358. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4359. in math_function.cpp, line 1227, tmp_value = 0
  4360. definition of type is half [FILL ZERO].
  4361. gspt r[i] = 0 (Q_frac=22
  4362. In blob.cpp, blob.Q_frac= 22
  4363. in math_function.cpp line 565, grp_num= 0, rand_index= 1
  4364. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  4365. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  4366. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4367. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  4368. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  4369. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  4370. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  4371. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4372. in math_function.cpp, line 730, tmp_value = 1835008
  4373. in math_function.cpp line 656, gspt_info_1[i]= 3092806, gspt_info_2[i]= 0
  4374. definition of type is half.
  4375. gspt r[i] = 0.0136719 (Q_frac=27
  4376. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4377. In blob.cpp, blob.Q_frac= 27
  4378. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4379. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4380. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4381. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4382. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4383. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4384. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4385. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4386. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4387. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4388. in math_function.cpp, line 1227, tmp_value = 0
  4389. definition of type is half [FILL ZERO].
  4390. gspt r[i] = 0 (Q_frac=22
  4391. In blob.cpp, blob.Q_frac= 22
  4392. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  4393. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  4394. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  4395. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  4396. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  4397. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  4398. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4399. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  4400. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4401. in math_function.cpp, line 730, tmp_value = -655360
  4402. in math_function.cpp line 656, gspt_info_1[i]= 13370058, gspt_info_2[i]= 0
  4403. definition of type is half.
  4404. gspt r[i] = -0.00488281 (Q_frac=27
  4405. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4406. In blob.cpp, blob.Q_frac= 27
  4407. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4408. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4409. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4410. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4411. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4412. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4413. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4414. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4415. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4416. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4417. in math_function.cpp, line 1227, tmp_value = 0
  4418. definition of type is half [FILL ZERO].
  4419. gspt r[i] = 0 (Q_frac=22
  4420. In blob.cpp, blob.Q_frac= 22
  4421. in math_function.cpp line 565, grp_num= 0, rand_index= 3
  4422. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  4423. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  4424. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  4425. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  4426. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  4427. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4428. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  4429. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4430. in math_function.cpp, line 730, tmp_value = 1966080
  4431. in math_function.cpp line 656, gspt_info_1[i]= 6931534, gspt_info_2[i]= 0
  4432. definition of type is half.
  4433. gspt r[i] = 0.0146484 (Q_frac=27
  4434. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4435. In blob.cpp, blob.Q_frac= 27
  4436. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4437. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4438. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4439. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4440. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4441. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4442. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4443. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4444. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4445. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4446. in math_function.cpp, line 1227, tmp_value = 0
  4447. definition of type is half [FILL ZERO].
  4448. gspt r[i] = 0 (Q_frac=22
  4449. In blob.cpp, blob.Q_frac= 22
  4450. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4451. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  4452. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  4453. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  4454. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  4455. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  4456. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4457. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  4458. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4459. in math_function.cpp, line 730, tmp_value = -917504
  4460. in math_function.cpp line 656, gspt_info_1[i]= 1188185, gspt_info_2[i]= 0
  4461. definition of type is half.
  4462. gspt r[i] = -0.00683594 (Q_frac=27
  4463. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4464. In blob.cpp, blob.Q_frac= 27
  4465. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4466. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4467. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4468. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4469. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4470. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4471. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4472. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4473. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4474. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4475. in math_function.cpp, line 1227, tmp_value = 0
  4476. definition of type is half [FILL ZERO].
  4477. gspt r[i] = 0 (Q_frac=22
  4478. In blob.cpp, blob.Q_frac= 22
  4479. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  4480. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  4481. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  4482. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  4483. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  4484. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  4485. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  4486. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  4487. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4488. in math_function.cpp, line 730, tmp_value = -2097152
  4489. in math_function.cpp line 656, gspt_info_1[i]= 14090448, gspt_info_2[i]= 0
  4490. definition of type is half.
  4491. gspt r[i] = -0.015625 (Q_frac=27
  4492. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4493. In blob.cpp, blob.Q_frac= 27
  4494. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4495. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4496. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4497. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4498. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4499. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4500. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4501. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4502. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4503. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4504. in math_function.cpp, line 1227, tmp_value = 0
  4505. definition of type is half [FILL ZERO].
  4506. gspt r[i] = 0 (Q_frac=22
  4507. In blob.cpp, blob.Q_frac= 22
  4508. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  4509. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  4510. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  4511. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4512. in math_function.cpp line 565, grp_num= 4, rand_index= 3
  4513. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  4514. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4515. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  4516. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4517. in math_function.cpp, line 730, tmp_value = -2228224
  4518. in math_function.cpp line 656, gspt_info_1[i]= 14030408, gspt_info_2[i]= 0
  4519. definition of type is half.
  4520. gspt r[i] = -0.00830078 (Q_frac=28
  4521. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4522. In blob.cpp, blob.Q_frac= 27
  4523. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4524. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4525. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4526. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4527. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4528. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4529. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4530. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4531. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4532. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4533. in math_function.cpp, line 1227, tmp_value = 0
  4534. definition of type is half [FILL ZERO].
  4535. gspt r[i] = 0 (Q_frac=22
  4536. In blob.cpp, blob.Q_frac= 22
  4537. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4538. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  4539. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  4540. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4541. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  4542. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  4543. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4544. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  4545. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4546. in math_function.cpp, line 730, tmp_value = 1966080
  4547. in math_function.cpp line 656, gspt_info_1[i]= 604686, gspt_info_2[i]= 0
  4548. definition of type is half.
  4549. gspt r[i] = 0.00732422 (Q_frac=28
  4550. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4551. In blob.cpp, blob.Q_frac= 27
  4552. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4553. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4554. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4555. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4556. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4557. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4558. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4559. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4560. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4561. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4562. in math_function.cpp, line 1227, tmp_value = 0
  4563. definition of type is half [FILL ZERO].
  4564. gspt r[i] = 0 (Q_frac=22
  4565. In blob.cpp, blob.Q_frac= 22
  4566. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  4567. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  4568. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  4569. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4570. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  4571. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  4572. in math_function.cpp line 565, grp_num= 6, rand_index= 0
  4573. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  4574. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4575. in math_function.cpp, line 730, tmp_value = 786432
  4576. in math_function.cpp line 656, gspt_info_1[i]= 11713093, gspt_info_2[i]= 0
  4577. definition of type is half.
  4578. gspt r[i] = 0.00292969 (Q_frac=28
  4579. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4580. In blob.cpp, blob.Q_frac= 27
  4581. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4582. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4583. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4584. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4585. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4586. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4587. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4588. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4589. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4590. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4591. in math_function.cpp, line 1227, tmp_value = 0
  4592. definition of type is half [FILL ZERO].
  4593. gspt r[i] = 0 (Q_frac=22
  4594. In blob.cpp, blob.Q_frac= 22
  4595. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  4596. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  4597. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  4598. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4599. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  4600. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  4601. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  4602. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  4603. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4604. in math_function.cpp, line 730, tmp_value = 2097152
  4605. in math_function.cpp line 656, gspt_info_1[i]= 9811990, gspt_info_2[i]= 0
  4606. definition of type is half.
  4607. gspt r[i] = 0.0078125 (Q_frac=28
  4608. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4609. In blob.cpp, blob.Q_frac= 27
  4610. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4611. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4612. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4613. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4614. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4615. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4616. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4617. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4618. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4619. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4620. in math_function.cpp, line 1227, tmp_value = 0
  4621. definition of type is half [FILL ZERO].
  4622. gspt r[i] = 0 (Q_frac=22
  4623. In blob.cpp, blob.Q_frac= 22
  4624. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  4625. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  4626. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  4627. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  4628. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  4629. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  4630. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  4631. in math_function.cpp line 565, grp_num= 7, rand_index= 2
  4632. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4633. in math_function.cpp, line 730, tmp_value = -262144
  4634. in math_function.cpp line 656, gspt_info_1[i]= 9544546, gspt_info_2[i]= 0
  4635. definition of type is half.
  4636. gspt r[i] = -0.000976562 (Q_frac=28
  4637. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4638. In blob.cpp, blob.Q_frac= 27
  4639. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4640. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4641. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4642. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4643. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4644. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4645. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4646. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4647. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4648. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4649. in math_function.cpp, line 1227, tmp_value = 0
  4650. definition of type is half [FILL ZERO].
  4651. gspt r[i] = 0 (Q_frac=22
  4652. In blob.cpp, blob.Q_frac= 22
  4653. in math_function.cpp line 565, grp_num= 0, rand_index= 1
  4654. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  4655. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  4656. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  4657. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  4658. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  4659. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4660. in math_function.cpp line 565, grp_num= 7, rand_index= 1
  4661. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4662. in math_function.cpp, line 730, tmp_value = -917504
  4663. in math_function.cpp line 656, gspt_info_1[i]= 2706137, gspt_info_2[i]= 0
  4664. definition of type is half.
  4665. gspt r[i] = -0.00341797 (Q_frac=28
  4666. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4667. In blob.cpp, blob.Q_frac= 28
  4668. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4669. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4670. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4671. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4672. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4673. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4674. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4675. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4676. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4677. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4678. in math_function.cpp, line 1227, tmp_value = 0
  4679. definition of type is half [FILL ZERO].
  4680. gspt r[i] = 0 (Q_frac=22
  4681. In blob.cpp, blob.Q_frac= 22
  4682. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  4683. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  4684. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  4685. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  4686. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  4687. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  4688. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4689. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  4690. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4691. in math_function.cpp, line 730, tmp_value = 1179648
  4692. in math_function.cpp line 656, gspt_info_1[i]= 10496733, gspt_info_2[i]= 0
  4693. definition of type is half.
  4694. gspt r[i] = 0.00439453 (Q_frac=28
  4695. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4696. In blob.cpp, blob.Q_frac= 28
  4697. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4698. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4699. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4700. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4701. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4702. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4703. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4704. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4705. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4706. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4707. in math_function.cpp, line 1227, tmp_value = 0
  4708. definition of type is half [FILL ZERO].
  4709. gspt r[i] = 0 (Q_frac=22
  4710. In blob.cpp, blob.Q_frac= 22
  4711. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4712. in math_function.cpp line 565, grp_num= 1, rand_index= 4
  4713. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  4714. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  4715. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  4716. in math_function.cpp line 565, grp_num= 5, rand_index= 1
  4717. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4718. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  4719. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4720. in math_function.cpp, line 730, tmp_value = -2228224
  4721. in math_function.cpp line 656, gspt_info_1[i]= 1125448, gspt_info_2[i]= 0
  4722. definition of type is half.
  4723. gspt r[i] = -0.00830078 (Q_frac=28
  4724. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4725. In blob.cpp, blob.Q_frac= 28
  4726. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4727. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4728. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4729. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4730. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4731. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4732. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4733. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4734. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4735. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4736. in math_function.cpp, line 1227, tmp_value = 0
  4737. definition of type is half [FILL ZERO].
  4738. gspt r[i] = 0 (Q_frac=22
  4739. In blob.cpp, blob.Q_frac= 22
  4740. in math_function.cpp line 565, grp_num= 0, rand_index= 6
  4741. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  4742. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  4743. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4744. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  4745. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  4746. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  4747. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  4748. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4749. in math_function.cpp, line 730, tmp_value = -2097152
  4750. in math_function.cpp line 656, gspt_info_1[i]= 14359888, gspt_info_2[i]= 0
  4751. definition of type is half.
  4752. gspt r[i] = -0.0078125 (Q_frac=28
  4753. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4754. In blob.cpp, blob.Q_frac= 28
  4755. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4756. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4757. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4758. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4759. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4760. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4761. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4762. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4763. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4764. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4765. in math_function.cpp, line 1227, tmp_value = 0
  4766. definition of type is half [FILL ZERO].
  4767. gspt r[i] = 0 (Q_frac=22
  4768. In blob.cpp, blob.Q_frac= 22
  4769. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4770. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  4771. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  4772. in math_function.cpp line 565, grp_num= 3, rand_index= 1
  4773. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  4774. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  4775. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4776. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  4777. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4778. in math_function.cpp, line 730, tmp_value = 1179648
  4779. in math_function.cpp line 656, gspt_info_1[i]= 956957, gspt_info_2[i]= 0
  4780. definition of type is half.
  4781. gspt r[i] = 0.00439453 (Q_frac=28
  4782. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4783. In blob.cpp, blob.Q_frac= 28
  4784. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4785. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4786. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4787. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4788. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4789. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4790. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4791. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4792. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4793. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4794. in math_function.cpp, line 1227, tmp_value = 0
  4795. definition of type is half [FILL ZERO].
  4796. gspt r[i] = 0 (Q_frac=22
  4797. In blob.cpp, blob.Q_frac= 22
  4798. in math_function.cpp line 565, grp_num= 0, rand_index= 2
  4799. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  4800. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  4801. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  4802. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  4803. in math_function.cpp line 565, grp_num= 5, rand_index= 4
  4804. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4805. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  4806. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4807. in math_function.cpp, line 730, tmp_value = 131072
  4808. in math_function.cpp line 656, gspt_info_1[i]= 5060891, gspt_info_2[i]= 0
  4809. definition of type is half.
  4810. gspt r[i] = 0.000488281 (Q_frac=28
  4811. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4812. In blob.cpp, blob.Q_frac= 28
  4813. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4814. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4815. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4816. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4817. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4818. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4819. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4820. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4821. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4822. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4823. in math_function.cpp, line 1227, tmp_value = 0
  4824. definition of type is half [FILL ZERO].
  4825. gspt r[i] = 0 (Q_frac=22
  4826. In blob.cpp, blob.Q_frac= 22
  4827. In blob.cpp, blob.Q_frac= 28
  4828. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4829. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4830. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4831. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4832. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4833. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4834. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4835. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4836. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4837. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4838. in math_function.cpp, line 1227, tmp_value = 0
  4839. definition of type is half [FILL ZERO].
  4840. gspt r[i] = 0 (Q_frac=22
  4841. In blob.cpp, blob.Q_frac= 22
  4842. in math_function.cpp line 565, grp_num= 0, rand_index= 5
  4843. in math_function.cpp line 565, grp_num= 1, rand_index= 3
  4844. in math_function.cpp line 565, grp_num= 2, rand_index= 5
  4845. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  4846. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  4847. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  4848. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  4849. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  4850. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4851. in math_function.cpp, line 730, tmp_value = 524288
  4852. in math_function.cpp line 656, gspt_info_1[i]= 11452564, gspt_info_2[i]= 0
  4853. definition of type is half.
  4854. gspt r[i] = 0.00195312 (Q_frac=28
  4855. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4856. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4857. in math_function.cpp line 565, grp_num= 1, rand_index= 6
  4858. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  4859. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  4860. in math_function.cpp line 565, grp_num= 4, rand_index= 5
  4861. in math_function.cpp line 565, grp_num= 5, rand_index= 6
  4862. in math_function.cpp line 565, grp_num= 6, rand_index= 1
  4863. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  4864. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4865. in math_function.cpp, line 730, tmp_value = 917504
  4866. in math_function.cpp line 656, gspt_info_1[i]= 1780621, gspt_info_2[i]= 0
  4867. definition of type is half.
  4868. gspt r[i] = 0.00341797 (Q_frac=28
  4869. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4870. In blob.cpp, blob.Q_frac= 28
  4871. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4872. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4873. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4874. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4875. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4876. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4877. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4878. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4879. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4880. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4881. in math_function.cpp, line 1227, tmp_value = 0
  4882. definition of type is half [FILL ZERO].
  4883. gspt r[i] = 0 (Q_frac=22
  4884. In blob.cpp, blob.Q_frac= 22
  4885. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  4886. in math_function.cpp line 565, grp_num= 1, rand_index= 2
  4887. in math_function.cpp line 565, grp_num= 2, rand_index= 6
  4888. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  4889. in math_function.cpp line 565, grp_num= 4, rand_index= 0
  4890. in math_function.cpp line 565, grp_num= 5, rand_index= 3
  4891. in math_function.cpp line 565, grp_num= 6, rand_index= 4
  4892. in math_function.cpp line 565, grp_num= 7, rand_index= 4
  4893. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4894. in math_function.cpp, line 730, tmp_value = 786432
  4895. in math_function.cpp line 656, gspt_info_1[i]= 9117924, gspt_info_2[i]= 0
  4896. definition of type is half.
  4897. gspt r[i] = 0.00585938 (Q_frac=27
  4898. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4899. In blob.cpp, blob.Q_frac= 27
  4900. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4901. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4902. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4903. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4904. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4905. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4906. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4907. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4908. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4909. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4910. in math_function.cpp, line 1227, tmp_value = 0
  4911. definition of type is half [FILL ZERO].
  4912. gspt r[i] = 0 (Q_frac=22
  4913. In blob.cpp, blob.Q_frac= 22
  4914. I0818 13:41:59.944496 18525 solver.cpp:55] Solver scaffolding done.
  4915. In blob.cpp, blob.Q_frac= 28
  4916. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4917. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4918. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4919. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4920. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4921. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4922. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4923. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4924. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4925. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4926. in math_function.cpp, line 1227, tmp_value = 0
  4927. definition of type is half [FILL ZERO].
  4928. gspt r[i] = 0 (Q_frac=22
  4929. In blob.cpp, blob.Q_frac= 22
  4930. in math_function.cpp line 565, grp_num= 0, rand_index= 3
  4931. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  4932. in math_function.cpp line 565, grp_num= 2, rand_index= 4
  4933. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  4934. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  4935. in math_function.cpp line 565, grp_num= 5, rand_index= 2
  4936. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  4937. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  4938. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4939. in math_function.cpp, line 730, tmp_value = 1048576
  4940. in math_function.cpp line 656, gspt_info_1[i]= 6431893, gspt_info_2[i]= 0
  4941. definition of type is half.
  4942. gspt r[i] = 0.0078125 (Q_frac=27
  4943. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4944. In blob.cpp, blob.Q_frac= 27
  4945. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4946. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4947. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4948. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4949. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4950. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4951. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4952. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4953. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4954. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4955. in math_function.cpp, line 1227, tmp_value = 0
  4956. definition of type is half [FILL ZERO].
  4957. gspt r[i] = 0 (Q_frac=22
  4958. In blob.cpp, blob.Q_frac= 22
  4959. I0818 13:42:23.051256 18528 solver.cpp:55] Solver scaffolding done.
  4960. In blob.cpp, blob.Q_frac= 28
  4961. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4962. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4963. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4964. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4965. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4966. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4967. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4968. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4969. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4970. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  4971. in math_function.cpp, line 1227, tmp_value = 0
  4972. definition of type is half [FILL ZERO].
  4973. gspt r[i] = 0 (Q_frac=22
  4974. In blob.cpp, blob.Q_frac= 22
  4975. in math_function.cpp line 565, grp_num= 0, rand_index= 0
  4976. in math_function.cpp line 565, grp_num= 1, rand_index= 1
  4977. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  4978. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  4979. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  4980. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  4981. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  4982. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  4983. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  4984. in math_function.cpp, line 730, tmp_value = 1179648
  4985. in math_function.cpp line 656, gspt_info_1[i]= 329053, gspt_info_2[i]= 0
  4986. definition of type is half.
  4987. gspt r[i] = 0.00878906 (Q_frac=27
  4988. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  4989. In blob.cpp, blob.Q_frac= 27
  4990. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  4991. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  4992. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  4993. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  4994. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  4995. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  4996. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  4997. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  4998. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  4999. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5000. in math_function.cpp, line 1227, tmp_value = 0
  5001. definition of type is half [FILL ZERO].
  5002. gspt r[i] = 0 (Q_frac=22
  5003. In blob.cpp, blob.Q_frac= 22
  5004. I0818 13:42:24.901659 18526 solver.cpp:55] Solver scaffolding done.
  5005. In blob.cpp, blob.Q_frac= 28
  5006. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  5007. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  5008. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  5009. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  5010. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  5011. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  5012. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  5013. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  5014. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  5015. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5016. in math_function.cpp, line 1227, tmp_value = 0
  5017. definition of type is half [FILL ZERO].
  5018. gspt r[i] = 0 (Q_frac=22
  5019. In blob.cpp, blob.Q_frac= 22
  5020. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  5021. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  5022. in math_function.cpp line 565, grp_num= 2, rand_index= 2
  5023. in math_function.cpp line 565, grp_num= 3, rand_index= 2
  5024. in math_function.cpp line 565, grp_num= 4, rand_index= 1
  5025. in math_function.cpp line 565, grp_num= 5, rand_index= 5
  5026. in math_function.cpp line 565, grp_num= 6, rand_index= 3
  5027. in math_function.cpp line 565, grp_num= 7, rand_index= 6
  5028. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  5029. in math_function.cpp, line 730, tmp_value = 2228224
  5030. in math_function.cpp line 656, gspt_info_1[i]= 8463198, gspt_info_2[i]= 0
  5031. definition of type is half.
  5032. gspt r[i] = 0.0166016 (Q_frac=27
  5033. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  5034. In blob.cpp, blob.Q_frac= 27
  5035. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  5036. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  5037. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  5038. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  5039. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  5040. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  5041. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  5042. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  5043. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  5044. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5045. in math_function.cpp, line 1227, tmp_value = 0
  5046. definition of type is half [FILL ZERO].
  5047. gspt r[i] = 0 (Q_frac=22
  5048. In blob.cpp, blob.Q_frac= 22
  5049. I0818 13:42:25.116623 18524 solver.cpp:55] Solver scaffolding done.
  5050. In blob.cpp, blob.Q_frac= 28
  5051. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  5052. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  5053. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  5054. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  5055. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  5056. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  5057. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  5058. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  5059. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  5060. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5061. in math_function.cpp, line 1227, tmp_value = 0
  5062. definition of type is half [FILL ZERO].
  5063. gspt r[i] = 0 (Q_frac=22
  5064. In blob.cpp, blob.Q_frac= 22
  5065. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  5066. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  5067. in math_function.cpp line 565, grp_num= 2, rand_index= 3
  5068. in math_function.cpp line 565, grp_num= 3, rand_index= 4
  5069. in math_function.cpp line 565, grp_num= 4, rand_index= 2
  5070. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  5071. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  5072. in math_function.cpp line 565, grp_num= 7, rand_index= 3
  5073. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  5074. in math_function.cpp, line 730, tmp_value = 0
  5075. in math_function.cpp line 656, gspt_info_1[i]= 9815059, gspt_info_2[i]= 0
  5076. definition of type is half.
  5077. gspt r[i] = 0 (Q_frac=27
  5078. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  5079. In blob.cpp, blob.Q_frac= 27
  5080. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  5081. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  5082. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  5083. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  5084. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  5085. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  5086. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  5087. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  5088. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  5089. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5090. in math_function.cpp, line 1227, tmp_value = 0
  5091. definition of type is half [FILL ZERO].
  5092. gspt r[i] = 0 (Q_frac=22
  5093. In blob.cpp, blob.Q_frac= 22
  5094. I0818 13:42:25.225204 18529 solver.cpp:55] Solver scaffolding done.
  5095. In blob.cpp, blob.Q_frac= 28
  5096. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  5097. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  5098. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  5099. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  5100. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  5101. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  5102. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  5103. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  5104. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  5105. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5106. in math_function.cpp, line 1227, tmp_value = 0
  5107. definition of type is half [FILL ZERO].
  5108. gspt r[i] = 0 (Q_frac=22
  5109. In blob.cpp, blob.Q_frac= 22
  5110. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  5111. in math_function.cpp line 565, grp_num= 1, rand_index= 0
  5112. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  5113. in math_function.cpp line 565, grp_num= 3, rand_index= 3
  5114. in math_function.cpp line 565, grp_num= 4, rand_index= 4
  5115. in math_function.cpp line 565, grp_num= 5, rand_index= 6
  5116. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  5117. in math_function.cpp line 565, grp_num= 7, rand_index= 0
  5118. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  5119. in math_function.cpp, line 730, tmp_value = -2097152
  5120. in math_function.cpp line 656, gspt_info_1[i]= 8403344, gspt_info_2[i]= 0
  5121. definition of type is half.
  5122. gspt r[i] = -0.015625 (Q_frac=27
  5123. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  5124. In blob.cpp, blob.Q_frac= 27
  5125. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  5126. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  5127. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  5128. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  5129. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  5130. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  5131. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  5132. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  5133. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  5134. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5135. in math_function.cpp, line 1227, tmp_value = 0
  5136. definition of type is half [FILL ZERO].
  5137. gspt r[i] = 0 (Q_frac=22
  5138. In blob.cpp, blob.Q_frac= 22
  5139. I0818 13:42:25.552917 18527 solver.cpp:55] Solver scaffolding done.
  5140. In blob.cpp, blob.Q_frac= 28
  5141. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  5142. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  5143. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  5144. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  5145. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  5146. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  5147. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  5148. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  5149. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  5150. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5151. in math_function.cpp, line 1227, tmp_value = 0
  5152. definition of type is half [FILL ZERO].
  5153. gspt r[i] = 0 (Q_frac=22
  5154. In blob.cpp, blob.Q_frac= 22
  5155. in math_function.cpp line 565, grp_num= 0, rand_index= 4
  5156. in math_function.cpp line 565, grp_num= 1, rand_index= 5
  5157. in math_function.cpp line 565, grp_num= 2, rand_index= 0
  5158. in math_function.cpp line 565, grp_num= 3, rand_index= 0
  5159. in math_function.cpp line 565, grp_num= 4, rand_index= 6
  5160. in math_function.cpp line 565, grp_num= 5, rand_index= 0
  5161. in math_function.cpp line 565, grp_num= 6, rand_index= 2
  5162. in math_function.cpp line 565, grp_num= 7, rand_index= 5
  5163. in math_function.cpp line 565, grp_num= 8, rand_index= 0
  5164. in math_function.cpp, line 730, tmp_value = 1048576
  5165. in math_function.cpp line 656, gspt_info_1[i]= 9702421, gspt_info_2[i]= 0
  5166. definition of type is half.
  5167. gspt r[i] = 0.0078125 (Q_frac=27
  5168. In fp32 2^(16):65536.00000000000000000000, in fp16 2^(16):inf
  5169. In blob.cpp, blob.Q_frac= 27
  5170. in math_function.cpp line 934, grp_num= 0, rand_index= 3
  5171. in math_function.cpp line 934, grp_num= 1, rand_index= 3
  5172. in math_function.cpp line 934, grp_num= 2, rand_index= 3
  5173. in math_function.cpp line 934, grp_num= 3, rand_index= 2
  5174. in math_function.cpp line 934, grp_num= 4, rand_index= 3
  5175. in math_function.cpp line 934, grp_num= 5, rand_index= 3
  5176. in math_function.cpp line 934, grp_num= 6, rand_index= 2
  5177. in math_function.cpp line 934, grp_num= 7, rand_index= 3
  5178. in math_function.cpp line 934, grp_num= 8, rand_index= 0
  5179. in math_function.cpp line 1223, gspt_info_1[i]= 7186131, gspt_info_2[i]= 0
  5180. in math_function.cpp, line 1227, tmp_value = 0
  5181. definition of type is half [FILL ZERO].
  5182. gspt r[i] = 0 (Q_frac=22
  5183. In blob.cpp, blob.Q_frac= 22
  5184. I0818 13:42:25.593865 18530 solver.cpp:55] Solver scaffolding done.
  5185. I0818 13:42:27.212508 18525 parallel.cpp:221] [2 - 2] P2PSync added callback
  5186. I0818 13:42:27.212574 18526 parallel.cpp:221] [3 - 3] P2PSync added callback
  5187. I0818 13:42:27.212579 18525 solver.cpp:418] [0.2] Solving CIFAR10_full_vgg7_BN Learning Rate Policy: multistep
  5188. I0818 13:42:27.212607 18526 solver.cpp:418] [0.3] Solving CIFAR10_full_vgg7_BN Learning Rate Policy: multistep
  5189. I0818 13:42:27.212648 18524 parallel.cpp:221] [1 - 1] P2PSync added callback
  5190. I0818 13:42:27.212641 18527 parallel.cpp:221] [4 - 4] P2PSync added callback
  5191. I0818 13:42:27.212641 18530 parallel.cpp:221] [7 - 7] P2PSync added callback
  5192. I0818 13:42:27.212682 18524 solver.cpp:418] [0.1] Solving CIFAR10_full_vgg7_BN Learning Rate Policy: multistep
  5193. I0818 13:42:27.212673 18528 parallel.cpp:221] [5 - 5] P2PSync added callback
  5194. I0818 13:42:27.212693 18529 parallel.cpp:221] [6 - 6] P2PSync added callback
  5195. I0818 13:42:27.212703 18523 parallel.cpp:221] [0 - 0] P2PSync added callback
  5196. I0818 13:42:27.212713 18530 solver.cpp:418] [0.7] Solving CIFAR10_full_vgg7_BN Learning Rate Policy: multistep
  5197. I0818 13:42:27.212711 18527 solver.cpp:418] [0.4] Solving CIFAR10_full_vgg7_BN Learning Rate Policy: multistep
  5198. I0818 13:42:27.212725 18528 solver.cpp:418] [0.5] Solving CIFAR10_full_vgg7_BN Learning Rate Policy: multistep
  5199. I0818 13:42:27.212725 18529 solver.cpp:418] [0.6] Solving CIFAR10_full_vgg7_BN Learning Rate Policy: multistep
  5200. I0818 13:42:27.212745 18523 solver.cpp:418] [0.0] Solving CIFAR10_full_vgg7_BN Learning Rate Policy: multistep
  5201. I0818 13:42:27.212940 18524 net.cpp:1441] [0.1] Reserving 25952640 bytes of shared learnable space for type FLOAT16
  5202. I0818 13:42:27.212947 18523 net.cpp:1441] [0.0] Reserving 25952640 bytes of shared learnable space for type FLOAT16
  5203. I0818 13:42:27.212941 18529 net.cpp:1441] [0.6] Reserving 25952640 bytes of shared learnable space for type FLOAT16
  5204. I0818 13:42:27.212949 18530 net.cpp:1441] [0.7] Reserving 25952640 bytes of shared learnable space for type FLOAT16
  5205. I0818 13:42:27.212945 18527 net.cpp:1441] [0.4] Reserving 25952640 bytes of shared learnable space for type FLOAT16
  5206. I0818 13:42:27.212947 18525 net.cpp:1441] [0.2] Reserving 25952640 bytes of shared learnable space for type FLOAT16
  5207. I0818 13:42:27.212949 18526 net.cpp:1441] [0.3] Reserving 25952640 bytes of shared learnable space for type FLOAT16
  5208. I0818 13:42:27.212941 18528 net.cpp:1441] [0.5] Reserving 25952640 bytes of shared learnable space for type FLOAT16
  5209. I0818 13:42:29.115339 18526 net.cpp:1441] [0.3] Reserving 2 bytes of shared learnable space for type FLOAT
  5210. I0818 13:42:29.127041 18523 net.cpp:1441] [0.0] Reserving 2 bytes of shared learnable space for type FLOAT
  5211. I0818 13:42:29.132092 18524 net.cpp:1441] [0.1] Reserving 2 bytes of shared learnable space for type FLOAT
  5212. I0818 13:42:29.135192 18525 net.cpp:1441] [0.2] Reserving 2 bytes of shared learnable space for type FLOAT
  5213. I0818 13:42:29.135278 18528 net.cpp:1441] [0.5] Reserving 2 bytes of shared learnable space for type FLOAT
  5214. I0818 13:42:29.138170 18527 net.cpp:1441] [0.4] Reserving 2 bytes of shared learnable space for type FLOAT
  5215. I0818 13:42:29.138799 18530 net.cpp:1441] [0.7] Reserving 2 bytes of shared learnable space for type FLOAT
  5216. I0818 13:42:29.140919 18529 net.cpp:1441] [0.6] Reserving 2 bytes of shared learnable space for type FLOAT
  5217. I0818 13:42:29.155326 18528 solver.cpp:220] [0.5] Starting Optimization on GPU 5
  5218. I0818 13:42:29.155330 18525 solver.cpp:220] [0.2] Starting Optimization on GPU 2
  5219. I0818 13:42:29.155323 18530 solver.cpp:220] [0.7] Starting Optimization on GPU 7
  5220. I0818 13:42:29.155319 18527 solver.cpp:220] [0.4] Starting Optimization on GPU 4
  5221. I0818 13:42:29.155334 18526 solver.cpp:220] [0.3] Starting Optimization on GPU 3
  5222. I0818 13:42:29.155336 18523 solver.cpp:220] [0.0] Starting Optimization on GPU 0
  5223. I0818 13:42:29.155334 18524 solver.cpp:220] [0.1] Starting Optimization on GPU 1
  5224. I0818 13:42:29.155326 18529 solver.cpp:220] [0.6] Starting Optimization on GPU 6
  5225. I0818 13:42:29.156385 19004 common.cpp:555] {4} NVML succeeded to set CPU affinity
  5226. I0818 13:42:29.157156 18523 solver.cpp:257] [MultiGPU] Initial Test started...
  5227. I0818 13:42:29.157287 18523 solver.cpp:501] Iteration 0, Testing net (#0)
  5228. I0818 13:42:29.157507 19005 common.cpp:555] {5} NVML succeeded to set CPU affinity
  5229. I0818 13:42:29.159587 19009 common.cpp:555] {1} NVML succeeded to set CPU affinity
  5230. I0818 13:42:29.159878 19007 common.cpp:555] {2} NVML succeeded to set CPU affinity
  5231. I0818 13:42:29.160125 19010 common.cpp:555] {0} NVML succeeded to set CPU affinity
  5232. I0818 13:42:29.160379 19008 common.cpp:555] {6} NVML succeeded to set CPU affinity
  5233. I0818 13:42:29.160756 19011 common.cpp:555] {3} NVML succeeded to set CPU affinity
  5234. I0818 13:42:29.160995 19006 common.cpp:555] {7} NVML succeeded to set CPU affinity
  5235. I0818 13:42:32.588143 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.0769043
  5236. I0818 13:42:32.588380 18523 solver.cpp:588] (0.0) Test net output #1: loss = 8.95312 (* 1 = 8.95312 loss)
  5237. I0818 13:42:32.588446 18523 solver.cpp:262] [MultiGPU] Initial Test completed in 3.43093s
  5238. I0818 13:42:32.603046 18559 internal_thread.cpp:42] Restarting 4 internal thread(s) on device 4
  5239. I0818 13:42:32.618099 18555 internal_thread.cpp:42] Restarting 4 internal thread(s) on device 2
  5240. I0818 13:42:32.632681 18560 internal_thread.cpp:42] Restarting 4 internal thread(s) on device 5
  5241. I0818 13:42:32.647644 18563 internal_thread.cpp:42] Restarting 4 internal thread(s) on device 6
  5242. I0818 13:42:32.663269 18500 internal_thread.cpp:42] Restarting 4 internal thread(s) on device 0
  5243. I0818 13:42:32.678661 18562 internal_thread.cpp:42] Restarting 4 internal thread(s) on device 1
  5244. I0818 13:42:32.695725 18565 internal_thread.cpp:42] Restarting 4 internal thread(s) on device 7
  5245. I0818 13:42:32.712205 18558 internal_thread.cpp:42] Restarting 4 internal thread(s) on device 3
  5246. I0818 13:42:32.712242 18559 internal_thread.cpp:18] Starting 1 internal thread(s) on device 4
  5247. I0818 13:42:32.712258 18555 internal_thread.cpp:18] Starting 1 internal thread(s) on device 2
  5248. I0818 13:42:32.712321 18560 internal_thread.cpp:18] Starting 1 internal thread(s) on device 5
  5249. I0818 13:42:32.712327 18563 internal_thread.cpp:18] Starting 1 internal thread(s) on device 6
  5250. I0818 13:42:32.712875 18562 internal_thread.cpp:18] Starting 1 internal thread(s) on device 1
  5251. I0818 13:42:32.712961 18559 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5252. I0818 13:42:32.712965 18555 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5253. I0818 13:42:32.713671 18500 internal_thread.cpp:18] Starting 1 internal thread(s) on device 0
  5254. I0818 13:42:32.714568 18563 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5255. I0818 13:42:32.714613 18562 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5256. I0818 13:42:32.714623 18565 internal_thread.cpp:18] Starting 1 internal thread(s) on device 7
  5257. I0818 13:42:32.716645 19051 common.cpp:555] {2} NVML succeeded to set CPU affinity
  5258. I0818 13:42:32.717833 18560 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5259. I0818 13:42:32.717934 19060 common.cpp:555] {7} NVML succeeded to set CPU affinity
  5260. I0818 13:42:32.719521 19052 common.cpp:555] {4} NVML succeeded to set CPU affinity
  5261. I0818 13:42:32.721210 19056 common.cpp:555] {5} NVML succeeded to set CPU affinity
  5262. I0818 13:42:32.722720 18500 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5263. I0818 13:42:32.722831 19053 common.cpp:555] {6} NVML succeeded to set CPU affinity
  5264. I0818 13:42:32.724376 18565 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5265. I0818 13:42:32.724660 19058 common.cpp:555] {0} NVML succeeded to set CPU affinity
  5266. I0818 13:42:32.727440 18558 internal_thread.cpp:18] Starting 1 internal thread(s) on device 3
  5267. I0818 13:42:32.727455 18555 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5268. I0818 13:42:32.727504 18559 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5269. I0818 13:42:32.727547 18563 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5270. I0818 13:42:32.727555 19054 common.cpp:555] {1} NVML succeeded to set CPU affinity
  5271. I0818 13:42:32.729012 18565 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5272. I0818 13:42:32.729038 18558 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5273. I0818 13:42:32.729089 18562 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5274. I0818 13:42:32.729115 18560 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5275. I0818 13:42:32.729125 19061 common.cpp:555] {3} NVML succeeded to set CPU affinity
  5276. I0818 13:42:32.729161 18500 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5277. I0818 13:42:32.729579 18559 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5278. I0818 13:42:32.729938 18555 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5279. I0818 13:42:32.730247 18563 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5280. I0818 13:42:32.734314 18555 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5281. I0818 13:42:32.734745 18562 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5282. I0818 13:42:32.735023 18558 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5283. I0818 13:42:32.735368 18560 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5284. I0818 13:42:32.735780 18500 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5285. I0818 13:42:32.736135 18559 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5286. I0818 13:42:32.736387 18565 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5287. I0818 13:42:32.736752 18563 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5288. I0818 13:42:32.740649 18555 data_reader.cpp:59] Data Reader threads: 3, out queues: 12, depth: 13
  5289. I0818 13:42:32.740773 18555 internal_thread.cpp:18] Starting 3 internal thread(s) on device 2
  5290. I0818 13:42:32.742493 19062 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5291. I0818 13:42:32.744063 19063 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5292. I0818 13:42:32.747342 18558 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5293. I0818 13:42:32.747368 19064 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5294. I0818 13:42:32.747408 18560 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5295. I0818 13:42:32.747457 18500 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5296. I0818 13:42:32.747481 18559 data_reader.cpp:59] Data Reader threads: 3, out queues: 12, depth: 13
  5297. I0818 13:42:32.747519 18562 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5298. I0818 13:42:32.747552 18565 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5299. I0818 13:42:32.747571 18563 data_reader.cpp:59] Data Reader threads: 3, out queues: 12, depth: 13
  5300. I0818 13:42:32.747615 18559 internal_thread.cpp:18] Starting 3 internal thread(s) on device 4
  5301. I0818 13:42:32.748270 18563 internal_thread.cpp:18] Starting 3 internal thread(s) on device 6
  5302. I0818 13:42:32.750777 19065 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5303. I0818 13:42:32.752236 19067 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5304. I0818 13:42:32.753692 19066 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5305. I0818 13:42:32.755179 19068 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5306. I0818 13:42:32.756763 19069 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5307. I0818 13:42:32.758507 18558 data_transformer.cpp:40] Loading mean file from: examples/cifar10/mean.binaryproto
  5308. I0818 13:42:32.758519 19070 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5309. I0818 13:42:32.758558 18560 data_reader.cpp:59] Data Reader threads: 3, out queues: 12, depth: 13
  5310. I0818 13:42:32.758606 18500 data_reader.cpp:59] Data Reader threads: 3, out queues: 12, depth: 13
  5311. I0818 13:42:32.758631 18562 data_reader.cpp:59] Data Reader threads: 3, out queues: 12, depth: 13
  5312. I0818 13:42:32.758687 18565 data_reader.cpp:59] Data Reader threads: 3, out queues: 12, depth: 13
  5313. I0818 13:42:32.758723 18560 internal_thread.cpp:18] Starting 3 internal thread(s) on device 5
  5314. I0818 13:42:32.758796 18500 internal_thread.cpp:18] Starting 3 internal thread(s) on device 0
  5315. I0818 13:42:32.758857 18562 internal_thread.cpp:18] Starting 3 internal thread(s) on device 1
  5316. I0818 13:42:32.759636 18565 internal_thread.cpp:18] Starting 3 internal thread(s) on device 7
  5317. I0818 13:42:32.763674 19072 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5318. I0818 13:42:32.764783 19073 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5319. I0818 13:42:32.766502 19075 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5320. I0818 13:42:32.767788 19071 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5321. I0818 13:42:32.769074 19074 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5322. I0818 13:42:32.770478 19079 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5323. I0818 13:42:32.771878 19077 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5324. I0818 13:42:32.776782 19076 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5325. I0818 13:42:32.778432 19078 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5326. I0818 13:42:32.779932 19080 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5327. I0818 13:42:32.781569 19081 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5328. I0818 13:42:32.783198 19082 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5329. I0818 13:42:32.783200 18558 data_reader.cpp:59] Data Reader threads: 3, out queues: 12, depth: 13
  5330. I0818 13:42:32.783341 18558 internal_thread.cpp:18] Starting 3 internal thread(s) on device 3
  5331. I0818 13:42:32.785162 19083 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5332. I0818 13:42:32.786742 19084 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5333. I0818 13:42:32.787921 18555 data_layer.cpp:200] [n0.d2.r2] Output data size: 13, 3, 28, 28
  5334. I0818 13:42:32.787962 18555 data_layer.cpp:105] [n0.d2.r2] Parser threads: 3 (auto)
  5335. I0818 13:42:32.787974 18555 data_layer.cpp:107] [n0.d2.r2] Transformer threads: 4 (auto)
  5336. I0818 13:42:32.787974 19085 db_lmdb.cpp:36] Opened lmdb examples/cifar10/cifar10_train_lmdb
  5337. I0818 13:42:32.830027 18559 data_layer.cpp:200] [n0.d4.r4] Output data size: 13, 3, 28, 28
  5338. I0818 13:42:32.830070 18559 data_layer.cpp:105] [n0.d4.r4] Parser threads: 3 (auto)
  5339. I0818 13:42:32.830080 18559 data_layer.cpp:107] [n0.d4.r4] Transformer threads: 4 (auto)
  5340. I0818 13:42:32.830909 18563 data_layer.cpp:200] [n0.d6.r6] Output data size: 13, 3, 28, 28
  5341. I0818 13:42:32.830971 18563 data_layer.cpp:105] [n0.d6.r6] Parser threads: 3 (auto)
  5342. I0818 13:42:32.831043 18563 data_layer.cpp:107] [n0.d6.r6] Transformer threads: 4 (auto)
  5343. I0818 13:42:32.850602 18560 data_layer.cpp:200] [n0.d5.r5] Output data size: 13, 3, 28, 28
  5344. I0818 13:42:32.850636 18560 data_layer.cpp:105] [n0.d5.r5] Parser threads: 3 (auto)
  5345. I0818 13:42:32.850644 18560 data_layer.cpp:107] [n0.d5.r5] Transformer threads: 4 (auto)
  5346. I0818 13:42:32.854048 18500 data_layer.cpp:200] [n0.d0.r0] Output data size: 13, 3, 28, 28
  5347. I0818 13:42:32.854082 18500 data_layer.cpp:105] [n0.d0.r0] Parser threads: 3 (auto)
  5348. I0818 13:42:32.854089 18500 data_layer.cpp:107] [n0.d0.r0] Transformer threads: 4 (auto)
  5349. I0818 13:42:32.854506 18562 data_layer.cpp:200] [n0.d1.r1] Output data size: 13, 3, 28, 28
  5350. I0818 13:42:32.854620 18562 data_layer.cpp:105] [n0.d1.r1] Parser threads: 3 (auto)
  5351. I0818 13:42:32.854629 18562 data_layer.cpp:107] [n0.d1.r1] Transformer threads: 4 (auto)
  5352. I0818 13:42:32.860416 18565 data_layer.cpp:200] [n0.d7.r7] Output data size: 13, 3, 28, 28
  5353. I0818 13:42:32.860455 18565 data_layer.cpp:105] [n0.d7.r7] Parser threads: 3 (auto)
  5354. I0818 13:42:32.860464 18565 data_layer.cpp:107] [n0.d7.r7] Transformer threads: 4 (auto)
  5355. I0818 13:42:32.882510 18558 data_layer.cpp:200] [n0.d3.r3] Output data size: 13, 3, 28, 28
  5356. I0818 13:42:32.882571 18558 data_layer.cpp:105] [n0.d3.r3] Parser threads: 3 (auto)
  5357. I0818 13:42:32.882583 18558 data_layer.cpp:107] [n0.d3.r3] Transformer threads: 4 (auto)
  5358. I0818 13:42:33.790786 18523 solver.cpp:341] [0.0] Iteration 0 (1.20218 s), loss = 2.31445
  5359. I0818 13:42:33.792801 18523 solver.cpp:361] [0.0] Train net output #0: loss = 2.31445 (* 1 = 2.31445 loss)
  5360. I0818 13:42:33.792850 18523 sgd_solver.cpp:180] [0.0] Iteration 0, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 1
  5361. I0818 13:42:34.165633 18523 solver.cpp:341] [0.0] Iteration 1 (0.374802 s), loss = 2.26367
  5362. I0818 13:42:34.165750 18523 solver.cpp:361] [0.0] Train net output #0: loss = 2.26367 (* 1 = 2.26367 loss)
  5363. I0818 13:42:34.330147 18529 cudnn_conv_layer.cpp:849] [n0.d6.r6] Conv Algos (F,BD,BF): 'conv1_1' with space 6.91K 3/1 1p 1 0p (avail 11.01G, req 6.91K) t: 0 0 0.11
  5364. I0818 13:42:34.340776 18530 cudnn_conv_layer.cpp:849] [n0.d7.r7] Conv Algos (F,BD,BF): 'conv1_1' with space 6.91K 3/1 1p 1 0p (avail 11.01G, req 6.91K) t: 0 0 0.12
  5365. I0818 13:42:34.360550 18528 cudnn_conv_layer.cpp:849] [n0.d5.r5] Conv Algos (F,BD,BF): 'conv1_1' with space 6.91K 3/1 1p 1 0p (avail 11.01G, req 6.91K) t: 0 0 0.12
  5366. I0818 13:42:34.365193 18524 cudnn_conv_layer.cpp:849] [n0.d1.r1] Conv Algos (F,BD,BF): 'conv1_1' with space 6.91K 3/1 1p 1 0p (avail 11.01G, req 6.91K) t: 0 0 0.11
  5367. I0818 13:42:34.368230 18525 cudnn_conv_layer.cpp:849] [n0.d2.r2] Conv Algos (F,BD,BF): 'conv1_1' with space 6.91K 3/1 1p 1 0p (avail 11.01G, req 6.91K) t: 0 0 0.12
  5368. I0818 13:42:34.371419 18526 cudnn_conv_layer.cpp:849] [n0.d3.r3] Conv Algos (F,BD,BF): 'conv1_1' with space 6.91K 3/1 1p 1 0p (avail 11.01G, req 6.91K) t: 0 0 0.12
  5369. I0818 13:42:34.374917 18523 cudnn_conv_layer.cpp:849] [n0.d0.r0] Conv Algos (F,BD,BF): 'conv1_1' with space 6.91K 3/1 1p 1 0p (avail 10.75G, req 6.91K) t: 0 0 0.13
  5370. I0818 13:42:34.378011 18527 cudnn_conv_layer.cpp:849] [n0.d4.r4] Conv Algos (F,BD,BF): 'conv1_1' with space 6.91K 3/1 1p 1 0p (avail 11.01G, req 6.91K) t: 0 0 0.12
  5371. I0818 13:42:34.759785 18530 cudnn_conv_layer.cpp:849] [n0.d7.r7] Conv Algos (F,BD,BF): 'conv1_2' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.23 0.23
  5372. I0818 13:42:34.771389 18525 cudnn_conv_layer.cpp:849] [n0.d2.r2] Conv Algos (F,BD,BF): 'conv1_2' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.23
  5373. I0818 13:42:34.947182 18529 cudnn_conv_layer.cpp:849] [n0.d6.r6] Conv Algos (F,BD,BF): 'conv1_2' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.22 0.23
  5374. I0818 13:42:35.012063 18528 cudnn_conv_layer.cpp:849] [n0.d5.r5] Conv Algos (F,BD,BF): 'conv1_2' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.23
  5375. I0818 13:42:35.042203 18523 cudnn_conv_layer.cpp:849] [n0.d0.r0] Conv Algos (F,BD,BF): 'conv1_2' with space 0.03G 128/1 6p 4p 5p (avail 10.72G, req 0.03G) t: 0 0.25 0.22
  5376. I0818 13:42:35.044123 18524 cudnn_conv_layer.cpp:849] [n0.d1.r1] Conv Algos (F,BD,BF): 'conv1_2' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.25 0.22
  5377. I0818 13:42:35.051460 18527 cudnn_conv_layer.cpp:849] [n0.d4.r4] Conv Algos (F,BD,BF): 'conv1_2' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.24
  5378. I0818 13:42:35.054173 18526 cudnn_conv_layer.cpp:849] [n0.d3.r3] Conv Algos (F,BD,BF): 'conv1_2' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.23
  5379. I0818 13:42:35.173856 18530 cudnn_conv_layer.cpp:849] [n0.d7.r7] Conv Algos (F,BD,BF): 'conv2_1' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.14 0.18
  5380. I0818 13:42:35.411159 18525 cudnn_conv_layer.cpp:849] [n0.d2.r2] Conv Algos (F,BD,BF): 'conv2_1' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.14 0.18
  5381. I0818 13:42:35.424628 18529 cudnn_conv_layer.cpp:849] [n0.d6.r6] Conv Algos (F,BD,BF): 'conv2_1' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.16 0.18
  5382. I0818 13:42:35.430181 18528 cudnn_conv_layer.cpp:849] [n0.d5.r5] Conv Algos (F,BD,BF): 'conv2_1' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.14 0.17
  5383. I0818 13:42:35.431936 18523 cudnn_conv_layer.cpp:849] [n0.d0.r0] Conv Algos (F,BD,BF): 'conv2_1' with space 0.03G 128/1 6p 4p 5p (avail 10.72G, req 0.03G) t: 0 0.14 0.17
  5384. I0818 13:42:35.436247 18526 cudnn_conv_layer.cpp:849] [n0.d3.r3] Conv Algos (F,BD,BF): 'conv2_1' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.13 0.18
  5385. I0818 13:42:35.484690 18527 cudnn_conv_layer.cpp:849] [n0.d4.r4] Conv Algos (F,BD,BF): 'conv2_1' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.15 0.19
  5386. I0818 13:42:35.599155 18524 cudnn_conv_layer.cpp:849] [n0.d1.r1] Conv Algos (F,BD,BF): 'conv2_1' with space 0.03G 128/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.14 0.18
  5387. I0818 13:42:35.740573 18530 cudnn_conv_layer.cpp:849] [n0.d7.r7] Conv Algos (F,BD,BF): 'conv2_2' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.26
  5388. I0818 13:42:35.982211 18525 cudnn_conv_layer.cpp:849] [n0.d2.r2] Conv Algos (F,BD,BF): 'conv2_2' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.25
  5389. I0818 13:42:35.996879 18529 cudnn_conv_layer.cpp:849] [n0.d6.r6] Conv Algos (F,BD,BF): 'conv2_2' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.25 0.27
  5390. I0818 13:42:35.998989 18523 cudnn_conv_layer.cpp:849] [n0.d0.r0] Conv Algos (F,BD,BF): 'conv2_2' with space 0.03G 256/1 6p 4p 5p (avail 10.72G, req 0.03G) t: 0 0.23 0.25
  5391. I0818 13:42:36.004915 18528 cudnn_conv_layer.cpp:849] [n0.d5.r5] Conv Algos (F,BD,BF): 'conv2_2' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.26
  5392. I0818 13:42:36.103113 18526 cudnn_conv_layer.cpp:849] [n0.d3.r3] Conv Algos (F,BD,BF): 'conv2_2' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.25
  5393. I0818 13:42:36.129307 18527 cudnn_conv_layer.cpp:849] [n0.d4.r4] Conv Algos (F,BD,BF): 'conv2_2' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.25 0.27
  5394. I0818 13:42:36.233860 18524 cudnn_conv_layer.cpp:849] [n0.d1.r1] Conv Algos (F,BD,BF): 'conv2_2' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.27
  5395. I0818 13:42:36.247576 18530 cudnn_conv_layer.cpp:849] [n0.d7.r7] Conv Algos (F,BD,BF): 'conv3_1' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.26
  5396. I0818 13:42:36.442348 18523 cudnn_conv_layer.cpp:849] [n0.d0.r0] Conv Algos (F,BD,BF): 'conv3_1' with space 0.03G 256/1 6p 4p 5p (avail 10.72G, req 0.03G) t: 0 0.24 0.26
  5397. I0818 13:42:36.451040 18525 cudnn_conv_layer.cpp:849] [n0.d2.r2] Conv Algos (F,BD,BF): 'conv3_1' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.25 0.25
  5398. I0818 13:42:36.572760 18529 cudnn_conv_layer.cpp:849] [n0.d6.r6] Conv Algos (F,BD,BF): 'conv3_1' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.25
  5399. I0818 13:42:36.661401 18527 cudnn_conv_layer.cpp:849] [n0.d4.r4] Conv Algos (F,BD,BF): 'conv3_1' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.25 0.26
  5400. I0818 13:42:36.668948 18526 cudnn_conv_layer.cpp:849] [n0.d3.r3] Conv Algos (F,BD,BF): 'conv3_1' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.24 0.26
  5401. I0818 13:42:36.682253 18528 cudnn_conv_layer.cpp:849] [n0.d5.r5] Conv Algos (F,BD,BF): 'conv3_1' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.25 0.26
  5402. I0818 13:42:36.684100 18524 cudnn_conv_layer.cpp:849] [n0.d1.r1] Conv Algos (F,BD,BF): 'conv3_1' with space 0.03G 256/1 6p 4p 5p (avail 10.99G, req 0.03G) t: 0 0.25 0.26
  5403. I0818 13:42:36.843667 18530 cudnn_conv_layer.cpp:849] [n0.d7.r7] Conv Algos (F,BD,BF): 'conv3_2' with space 0.05G 512/1 6p 5p 5p (avail 10.97G, req 0.05G) t: 0 0.45 0.41
  5404. I0818 13:42:36.994576 18523 cudnn_conv_layer.cpp:849] [n0.d0.r0] Conv Algos (F,BD,BF): 'conv3_2' with space 0.05G 512/1 6p 5p 5p (avail 10.7G, req 0.05G) t: 0 0.45 0.39
  5405. I0818 13:42:37.094976 18525 cudnn_conv_layer.cpp:849] [n0.d2.r2] Conv Algos (F,BD,BF): 'conv3_2' with space 0.05G 512/1 6p 5p 5p (avail 10.97G, req 0.05G) t: 0 0.45 0.42
  5406. I0818 13:42:37.145619 18527 cudnn_conv_layer.cpp:849] [n0.d4.r4] Conv Algos (F,BD,BF): 'conv3_2' with space 0.05G 512/1 6p 5p 5p (avail 10.97G, req 0.05G) t: 0 0.44 0.41
  5407. I0818 13:42:37.156098 18529 cudnn_conv_layer.cpp:849] [n0.d6.r6] Conv Algos (F,BD,BF): 'conv3_2' with space 0.05G 512/1 6p 5p 5p (avail 10.97G, req 0.05G) t: 0 0.45 0.41
  5408. I0818 13:42:37.210855 18526 cudnn_conv_layer.cpp:849] [n0.d3.r3] Conv Algos (F,BD,BF): 'conv3_2' with space 0.05G 512/1 6p 5p 5p (avail 10.97G, req 0.05G) t: 0 0.45 0.41
  5409. I0818 13:42:37.251657 18528 cudnn_conv_layer.cpp:849] [n0.d5.r5] Conv Algos (F,BD,BF): 'conv3_2' with space 0.05G 512/1 6p 5p 5p (avail 10.97G, req 0.05G) t: 0 0.44 0.42
  5410. I0818 13:42:37.256307 18524 cudnn_conv_layer.cpp:849] [n0.d1.r1] Conv Algos (F,BD,BF): 'conv3_2' with space 0.05G 512/1 6p 5p 5p (avail 10.97G, req 0.05G) t: 0 0.44 0.4
  5411. I0818 13:42:37.366358 18523 solver.cpp:341] [0.0] Iteration 2 (3.20051 s), loss = 2.28516
  5412. I0818 13:42:37.366591 18523 solver.cpp:361] [0.0] Train net output #0: loss = 2.28516 (* 1 = 2.28516 loss)
  5413. I0818 13:42:59.966665 18523 solver.cpp:333] [0.0] Iteration 200 (8.76146 iter/s, 22.599s/198 iter), loss = 1.5293
  5414. I0818 13:42:59.966783 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.5293 (* 1 = 1.5293 loss)
  5415. I0818 13:42:59.966806 18523 sgd_solver.cpp:180] [0.0] Iteration 200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5416. I0818 13:43:22.788758 18523 solver.cpp:333] [0.0] Iteration 400 (8.76398 iter/s, 22.8207s/200 iter), loss = 1.48145
  5417. I0818 13:43:22.789324 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.48145 (* 1 = 1.48145 loss)
  5418. I0818 13:43:22.789343 18523 sgd_solver.cpp:180] [0.0] Iteration 400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5419. I0818 13:43:28.922415 19072 data_reader.cpp:321] Restarting data pre-fetching
  5420. I0818 13:43:46.146026 18523 solver.cpp:333] [0.0] Iteration 600 (8.56312 iter/s, 23.356s/200 iter), 1.2/104ep, loss = 1.34082
  5421. I0818 13:43:46.146168 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.34082 (* 1 = 1.34082 loss)
  5422. I0818 13:43:46.146201 18523 sgd_solver.cpp:180] [0.0] Iteration 600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5423. I0818 13:44:10.194103 18523 solver.cpp:333] [0.0] Iteration 800 (8.31716 iter/s, 24.0467s/200 iter), 1.7/104ep, loss = 1.57617
  5424. I0818 13:44:10.195009 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.57617 (* 1 = 1.57617 loss)
  5425. I0818 13:44:10.195029 18523 sgd_solver.cpp:180] [0.0] Iteration 800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5426. I0818 13:44:26.664268 19072 data_reader.cpp:321] Restarting data pre-fetching
  5427. I0818 13:44:34.812685 18523 solver.cpp:501] Iteration 1000, Testing net (#0)
  5428. I0818 13:44:43.081737 18511 data_reader.cpp:321] Restarting data pre-fetching
  5429. I0818 13:44:44.415019 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.418384
  5430. I0818 13:44:44.415061 18523 solver.cpp:588] (0.0) Test net output #1: loss = 1.78435 (* 1 = 1.78435 loss)
  5431. I0818 13:44:44.415189 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.2191s
  5432. I0818 13:44:44.536296 18523 solver.cpp:333] [0.0] Iteration 1000 (5.84469 iter/s, 34.2191s/200 iter), 2.1/104ep, loss = 1.85254
  5433. I0818 13:44:44.536484 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.85254 (* 1 = 1.85254 loss)
  5434. I0818 13:44:44.536546 18523 sgd_solver.cpp:180] [0.0] Iteration 1000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5435. I0818 13:45:09.314909 18523 solver.cpp:333] [0.0] Iteration 1200 (8.07195 iter/s, 24.7772s/200 iter), 2.5/104ep, loss = 1.36328
  5436. I0818 13:45:09.315044 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.36328 (* 1 = 1.36328 loss)
  5437. I0818 13:45:09.315114 18523 sgd_solver.cpp:180] [0.0] Iteration 1200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5438. I0818 13:45:34.405330 18523 solver.cpp:333] [0.0] Iteration 1400 (7.97163 iter/s, 25.089s/200 iter), 2.9/104ep, loss = 0.914062
  5439. I0818 13:45:34.405634 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.914062 (* 1 = 0.914062 loss)
  5440. I0818 13:45:34.405658 18523 sgd_solver.cpp:180] [0.0] Iteration 1400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5441. I0818 13:45:36.046392 19072 data_reader.cpp:321] Restarting data pre-fetching
  5442. I0818 13:45:59.398661 18523 solver.cpp:333] [0.0] Iteration 1600 (8.00257 iter/s, 24.992s/200 iter), 3.3/104ep, loss = 0.636719
  5443. I0818 13:45:59.398788 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.636719 (* 1 = 0.636719 loss)
  5444. I0818 13:45:59.398810 18523 sgd_solver.cpp:180] [0.0] Iteration 1600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5445. I0818 13:46:24.499106 18523 solver.cpp:333] [0.0] Iteration 1800 (7.96843 iter/s, 25.099s/200 iter), 3.7/104ep, loss = 1.13477
  5446. I0818 13:46:24.499994 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.13477 (* 1 = 1.13477 loss)
  5447. I0818 13:46:24.500010 18523 sgd_solver.cpp:180] [0.0] Iteration 1800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5448. I0818 13:46:36.528798 19072 data_reader.cpp:321] Restarting data pre-fetching
  5449. I0818 13:46:49.543622 18523 solver.cpp:501] Iteration 2000, Testing net (#0)
  5450. I0818 13:46:57.501245 18511 data_reader.cpp:321] Restarting data pre-fetching
  5451. I0818 13:46:58.904381 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.588411
  5452. I0818 13:46:58.904444 18523 solver.cpp:588] (0.0) Test net output #1: loss = 1.22713 (* 1 = 1.22713 loss)
  5453. I0818 13:46:58.904539 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.4036s
  5454. I0818 13:46:59.023461 18523 solver.cpp:333] [0.0] Iteration 2000 (5.81334 iter/s, 34.4036s/200 iter), 4.2/104ep, loss = 0.907715
  5455. I0818 13:46:59.023691 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.907715 (* 1 = 0.907715 loss)
  5456. I0818 13:46:59.023718 18523 sgd_solver.cpp:180] [0.0] Iteration 2000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5457. I0818 13:47:23.846411 18523 solver.cpp:333] [0.0] Iteration 2200 (8.05749 iter/s, 24.8216s/200 iter), 4.6/104ep, loss = 1.2002
  5458. I0818 13:47:23.846576 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.2002 (* 1 = 1.2002 loss)
  5459. I0818 13:47:23.846617 18523 sgd_solver.cpp:180] [0.0] Iteration 2200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5460. I0818 13:47:45.719180 19072 data_reader.cpp:321] Restarting data pre-fetching
  5461. I0818 13:47:48.767408 18523 solver.cpp:333] [0.0] Iteration 2400 (8.02581 iter/s, 24.9196s/200 iter), 5/104ep, loss = 1.11914
  5462. I0818 13:47:48.767592 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.11914 (* 1 = 1.11914 loss)
  5463. I0818 13:47:48.767621 18523 sgd_solver.cpp:180] [0.0] Iteration 2400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5464. I0818 13:48:13.750476 18523 solver.cpp:333] [0.0] Iteration 2600 (8.00585 iter/s, 24.9817s/200 iter), 5.4/104ep, loss = 1.43555
  5465. I0818 13:48:13.750682 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.43555 (* 1 = 1.43555 loss)
  5466. I0818 13:48:13.750726 18523 sgd_solver.cpp:180] [0.0] Iteration 2600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5467. I0818 13:48:38.911048 18523 solver.cpp:333] [0.0] Iteration 2800 (7.94938 iter/s, 25.1592s/200 iter), 5.8/104ep, loss = 1.125
  5468. I0818 13:48:38.911515 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.125 (* 1 = 1.125 loss)
  5469. I0818 13:48:38.911563 18523 sgd_solver.cpp:180] [0.0] Iteration 2800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5470. I0818 13:48:46.028232 19072 data_reader.cpp:321] Restarting data pre-fetching
  5471. I0818 13:49:03.936349 18523 solver.cpp:501] Iteration 3000, Testing net (#0)
  5472. I0818 13:49:11.584172 18511 data_reader.cpp:321] Restarting data pre-fetching
  5473. I0818 13:49:13.295364 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.59227
  5474. I0818 13:49:13.295400 18523 solver.cpp:588] (0.0) Test net output #1: loss = 1.26887 (* 1 = 1.26887 loss)
  5475. I0818 13:49:13.295434 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.3825s
  5476. I0818 13:49:13.417239 18523 solver.cpp:333] [0.0] Iteration 3000 (5.81691 iter/s, 34.3825s/200 iter), 6.2/104ep, loss = 0.714355
  5477. I0818 13:49:13.417347 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.714355 (* 1 = 0.714355 loss)
  5478. I0818 13:49:13.417392 18523 sgd_solver.cpp:180] [0.0] Iteration 3000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5479. I0818 13:49:38.289185 18523 solver.cpp:333] [0.0] Iteration 3200 (8.04163 iter/s, 24.8706s/200 iter), 6.7/104ep, loss = 0.974121
  5480. I0818 13:49:38.289326 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.974121 (* 1 = 0.974121 loss)
  5481. I0818 13:49:38.289366 18523 sgd_solver.cpp:180] [0.0] Iteration 3200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5482. I0818 13:49:55.473922 19072 data_reader.cpp:321] Restarting data pre-fetching
  5483. I0818 13:50:03.558075 18523 solver.cpp:333] [0.0] Iteration 3400 (7.9153 iter/s, 25.2675s/200 iter), 7.1/104ep, loss = 0.945312
  5484. I0818 13:50:03.558270 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.945312 (* 1 = 0.945312 loss)
  5485. I0818 13:50:03.558300 18523 sgd_solver.cpp:180] [0.0] Iteration 3400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5486. I0818 13:50:28.501219 18523 solver.cpp:333] [0.0] Iteration 3600 (8.0187 iter/s, 24.9417s/200 iter), 7.5/104ep, loss = 0.950684
  5487. I0818 13:50:28.501462 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.950684 (* 1 = 0.950684 loss)
  5488. I0818 13:50:28.501479 18523 sgd_solver.cpp:180] [0.0] Iteration 3600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5489. I0818 13:50:53.521579 18523 solver.cpp:333] [0.0] Iteration 3800 (7.99401 iter/s, 25.0187s/200 iter), 7.9/104ep, loss = 0.838867
  5490. I0818 13:50:53.521697 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.838867 (* 1 = 0.838867 loss)
  5491. I0818 13:50:53.521720 18523 sgd_solver.cpp:180] [0.0] Iteration 3800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5492. I0818 13:50:55.853032 19072 data_reader.cpp:321] Restarting data pre-fetching
  5493. I0818 13:51:18.719725 18523 solver.cpp:501] Iteration 4000, Testing net (#0)
  5494. I0818 13:51:25.932190 18511 data_reader.cpp:321] Restarting data pre-fetching
  5495. I0818 13:51:28.342186 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.641489
  5496. I0818 13:51:28.342223 18523 solver.cpp:588] (0.0) Test net output #1: loss = 1.0017 (* 1 = 1.0017 loss)
  5497. I0818 13:51:28.342312 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.8186s
  5498. I0818 13:51:28.465741 18523 solver.cpp:333] [0.0] Iteration 4000 (5.74406 iter/s, 34.8186s/200 iter), 8.3/104ep, loss = 0.417236
  5499. I0818 13:51:28.465836 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.417236 (* 1 = 0.417236 loss)
  5500. I0818 13:51:28.465858 18523 sgd_solver.cpp:180] [0.0] Iteration 4000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5501. I0818 13:51:53.407313 18523 solver.cpp:333] [0.0] Iteration 4200 (8.01924 iter/s, 24.94s/200 iter), 8.7/104ep, loss = 0.115356
  5502. I0818 13:51:53.407552 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.115356 (* 1 = 0.115356 loss)
  5503. I0818 13:51:53.407599 18523 sgd_solver.cpp:180] [0.0] Iteration 4200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5504. I0818 13:52:06.074932 19072 data_reader.cpp:321] Restarting data pre-fetching
  5505. I0818 13:52:18.884608 18523 solver.cpp:333] [0.0] Iteration 4400 (7.85062 iter/s, 25.4757s/200 iter), 9.2/104ep, loss = 0.931152
  5506. I0818 13:52:18.884683 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.931152 (* 1 = 0.931152 loss)
  5507. I0818 13:52:18.884706 18523 sgd_solver.cpp:180] [0.0] Iteration 4400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5508. I0818 13:52:44.005352 18523 solver.cpp:333] [0.0] Iteration 4600 (7.96205 iter/s, 25.1192s/200 iter), 9.6/104ep, loss = 1.44629
  5509. I0818 13:52:44.005563 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.44629 (* 1 = 1.44629 loss)
  5510. I0818 13:52:44.005586 18523 sgd_solver.cpp:180] [0.0] Iteration 4600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5511. I0818 13:53:06.493311 19072 data_reader.cpp:321] Restarting data pre-fetching
  5512. I0818 13:53:09.176096 18523 solver.cpp:333] [0.0] Iteration 4800 (7.94622 iter/s, 25.1692s/200 iter), 10/104ep, loss = 1.08105
  5513. I0818 13:53:09.176285 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.08105 (* 1 = 1.08105 loss)
  5514. I0818 13:53:09.176340 18523 sgd_solver.cpp:180] [0.0] Iteration 4800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5515. I0818 13:53:34.246821 18523 solver.cpp:501] Iteration 5000, Testing net (#0)
  5516. I0818 13:53:41.077462 18511 data_reader.cpp:321] Restarting data pre-fetching
  5517. I0818 13:53:43.665791 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.734543
  5518. I0818 13:53:43.665828 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.799883 (* 1 = 0.799883 loss)
  5519. I0818 13:53:43.665897 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.4877s
  5520. I0818 13:53:43.786623 18523 solver.cpp:333] [0.0] Iteration 5000 (5.79916 iter/s, 34.4877s/200 iter), 10.4/104ep, loss = 0.793945
  5521. I0818 13:53:43.786785 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.793945 (* 1 = 0.793945 loss)
  5522. I0818 13:53:43.786828 18523 sgd_solver.cpp:180] [0.0] Iteration 5000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5523. I0818 13:54:08.522706 18523 solver.cpp:333] [0.0] Iteration 5200 (8.08586 iter/s, 24.7345s/200 iter), 10.8/104ep, loss = 0.477539
  5524. I0818 13:54:08.522948 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.477539 (* 1 = 0.477539 loss)
  5525. I0818 13:54:08.522974 18523 sgd_solver.cpp:180] [0.0] Iteration 5200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5526. I0818 13:54:16.221410 19072 data_reader.cpp:321] Restarting data pre-fetching
  5527. I0818 13:54:34.124281 18523 solver.cpp:333] [0.0] Iteration 5400 (7.81249 iter/s, 25.6s/200 iter), 11.2/104ep, loss = 0.787109
  5528. I0818 13:54:34.124430 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.787109 (* 1 = 0.787109 loss)
  5529. I0818 13:54:34.124482 18523 sgd_solver.cpp:180] [0.0] Iteration 5400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5530. I0818 13:54:59.196501 18523 solver.cpp:333] [0.0] Iteration 5600 (7.97745 iter/s, 25.0707s/200 iter), 11.6/104ep, loss = 1.43164
  5531. I0818 13:54:59.196851 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.43164 (* 1 = 1.43164 loss)
  5532. I0818 13:54:59.196872 18523 sgd_solver.cpp:180] [0.0] Iteration 5600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5533. I0818 13:55:17.095489 19072 data_reader.cpp:321] Restarting data pre-fetching
  5534. I0818 13:55:24.457403 18523 solver.cpp:333] [0.0] Iteration 5800 (7.91785 iter/s, 25.2594s/200 iter), 12.1/104ep, loss = 0.745605
  5535. I0818 13:55:24.457662 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.745605 (* 1 = 0.745605 loss)
  5536. I0818 13:55:24.457751 18523 sgd_solver.cpp:180] [0.0] Iteration 5800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5537. I0818 13:55:49.366029 18523 solver.cpp:501] Iteration 6000, Testing net (#0)
  5538. I0818 13:55:55.847252 18511 data_reader.cpp:321] Restarting data pre-fetching
  5539. I0818 13:55:58.760283 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.693804
  5540. I0818 13:55:58.760318 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.875842 (* 1 = 0.875842 loss)
  5541. I0818 13:55:58.760388 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.301s
  5542. I0818 13:55:58.880888 18523 solver.cpp:333] [0.0] Iteration 6000 (5.83074 iter/s, 34.301s/200 iter), 12.5/104ep, loss = 0.679199
  5543. I0818 13:55:58.880986 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.679199 (* 1 = 0.679199 loss)
  5544. I0818 13:55:58.881011 18523 sgd_solver.cpp:180] [0.0] Iteration 6000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5545. I0818 13:56:23.877193 18523 solver.cpp:333] [0.0] Iteration 6200 (8.00166 iter/s, 24.9948s/200 iter), 12.9/104ep, loss = 0.453125
  5546. I0818 13:56:23.877547 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.453125 (* 1 = 0.453125 loss)
  5547. I0818 13:56:23.877588 18523 sgd_solver.cpp:180] [0.0] Iteration 6200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5548. I0818 13:56:26.605276 19072 data_reader.cpp:321] Restarting data pre-fetching
  5549. I0818 13:56:48.975360 18523 solver.cpp:333] [0.0] Iteration 6400 (7.96918 iter/s, 25.0967s/200 iter), 13.3/104ep, loss = 0.326416
  5550. I0818 13:56:48.975476 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.326416 (* 1 = 0.326416 loss)
  5551. I0818 13:56:48.975498 18523 sgd_solver.cpp:180] [0.0] Iteration 6400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5552. I0818 13:57:14.092380 18523 solver.cpp:333] [0.0] Iteration 6600 (7.9632 iter/s, 25.1155s/200 iter), 13.7/104ep, loss = 0.467773
  5553. I0818 13:57:14.093147 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.467773 (* 1 = 0.467773 loss)
  5554. I0818 13:57:14.093199 18523 sgd_solver.cpp:180] [0.0] Iteration 6600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5555. I0818 13:57:27.016713 19072 data_reader.cpp:321] Restarting data pre-fetching
  5556. I0818 13:57:39.494773 18523 solver.cpp:333] [0.0] Iteration 6800 (7.87376 iter/s, 25.4008s/200 iter), 14.1/104ep, loss = 0.487793
  5557. I0818 13:57:39.494974 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.487793 (* 1 = 0.487793 loss)
  5558. I0818 13:57:39.495038 18523 sgd_solver.cpp:180] [0.0] Iteration 6800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5559. I0818 13:58:04.943470 18523 solver.cpp:501] Iteration 7000, Testing net (#0)
  5560. I0818 13:58:11.146018 18511 data_reader.cpp:321] Restarting data pre-fetching
  5561. I0818 13:58:14.659453 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.733792
  5562. I0818 13:58:14.659510 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.761086 (* 1 = 0.761086 loss)
  5563. I0818 13:58:14.659603 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.1628s
  5564. I0818 13:58:14.786521 18523 solver.cpp:333] [0.0] Iteration 7000 (5.68782 iter/s, 35.1628s/200 iter), 14.6/104ep, loss = 0.410645
  5565. I0818 13:58:14.786669 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.410645 (* 1 = 0.410645 loss)
  5566. I0818 13:58:14.786725 18523 sgd_solver.cpp:180] [0.0] Iteration 7000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5567. I0818 13:58:37.964666 19072 data_reader.cpp:321] Restarting data pre-fetching
  5568. I0818 13:58:40.244642 18523 solver.cpp:333] [0.0] Iteration 7200 (7.8565 iter/s, 25.4566s/200 iter), 15/104ep, loss = 0.134155
  5569. I0818 13:58:40.244758 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.134155 (* 1 = 0.134155 loss)
  5570. I0818 13:58:40.244814 18523 sgd_solver.cpp:180] [0.0] Iteration 7200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5571. I0818 13:59:05.899322 18523 solver.cpp:333] [0.0] Iteration 7400 (7.79631 iter/s, 25.6532s/200 iter), 15.4/104ep, loss = 0.669922
  5572. I0818 13:59:05.899437 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.669922 (* 1 = 0.669922 loss)
  5573. I0818 13:59:05.899454 18523 sgd_solver.cpp:180] [0.0] Iteration 7400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5574. I0818 13:59:31.820060 18523 solver.cpp:333] [0.0] Iteration 7600 (7.71629 iter/s, 25.9192s/200 iter), 15.8/104ep, loss = 0.387207
  5575. I0818 13:59:31.820581 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.387207 (* 1 = 0.387207 loss)
  5576. I0818 13:59:31.820677 18523 sgd_solver.cpp:180] [0.0] Iteration 7600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5577. I0818 13:59:40.119488 19072 data_reader.cpp:321] Restarting data pre-fetching
  5578. I0818 13:59:57.402424 18523 solver.cpp:333] [0.0] Iteration 7800 (7.81833 iter/s, 25.5809s/200 iter), 16.2/104ep, loss = 0.497559
  5579. I0818 13:59:57.402513 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.497559 (* 1 = 0.497559 loss)
  5580. I0818 13:59:57.402536 18523 sgd_solver.cpp:180] [0.0] Iteration 7800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5581. I0818 14:00:22.849087 18523 solver.cpp:501] Iteration 8000, Testing net (#0)
  5582. I0818 14:00:28.682356 18511 data_reader.cpp:321] Restarting data pre-fetching
  5583. I0818 14:00:32.454154 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.76301
  5584. I0818 14:00:32.454211 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.665333 (* 1 = 0.665333 loss)
  5585. I0818 14:00:32.454294 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.0499s
  5586. I0818 14:00:32.577229 18523 solver.cpp:333] [0.0] Iteration 8000 (5.70615 iter/s, 35.0499s/200 iter), 16.6/104ep, loss = 0.247803
  5587. I0818 14:00:32.577574 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.247803 (* 1 = 0.247803 loss)
  5588. I0818 14:00:32.577612 18523 sgd_solver.cpp:180] [0.0] Iteration 8000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5589. I0818 14:00:50.929989 19072 data_reader.cpp:321] Restarting data pre-fetching
  5590. I0818 14:00:57.949189 18523 solver.cpp:333] [0.0] Iteration 8200 (7.88318 iter/s, 25.3705s/200 iter), 17.1/104ep, loss = 0.438232
  5591. I0818 14:00:57.949674 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.438232 (* 1 = 0.438232 loss)
  5592. I0818 14:00:57.949714 18523 sgd_solver.cpp:180] [0.0] Iteration 8200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5593. I0818 14:01:23.590920 18523 solver.cpp:333] [0.0] Iteration 8400 (7.80024 iter/s, 25.6402s/200 iter), 17.5/104ep, loss = 1.07422
  5594. I0818 14:01:23.591015 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.07422 (* 1 = 1.07422 loss)
  5595. I0818 14:01:23.591029 18523 sgd_solver.cpp:180] [0.0] Iteration 8400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5596. I0818 14:01:49.090409 18523 solver.cpp:333] [0.0] Iteration 8600 (7.84374 iter/s, 25.498s/200 iter), 17.9/104ep, loss = 1.28906
  5597. I0818 14:01:49.091529 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.28906 (* 1 = 1.28906 loss)
  5598. I0818 14:01:49.091548 18523 sgd_solver.cpp:180] [0.0] Iteration 8600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5599. I0818 14:01:52.254532 19072 data_reader.cpp:321] Restarting data pre-fetching
  5600. I0818 14:02:14.377588 18523 solver.cpp:333] [0.0] Iteration 8800 (7.90961 iter/s, 25.2857s/200 iter), 18.3/104ep, loss = 0.741699
  5601. I0818 14:02:14.377748 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.741699 (* 1 = 0.741699 loss)
  5602. I0818 14:02:14.377800 18523 sgd_solver.cpp:180] [0.0] Iteration 8800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5603. I0818 14:02:40.169919 18523 solver.cpp:501] Iteration 9000, Testing net (#0)
  5604. I0818 14:02:45.658071 18511 data_reader.cpp:321] Restarting data pre-fetching
  5605. I0818 14:02:49.677757 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.756855
  5606. I0818 14:02:49.677809 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.707629 (* 1 = 0.707629 loss)
  5607. I0818 14:02:49.677917 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.2984s
  5608. I0818 14:02:49.803516 18523 solver.cpp:333] [0.0] Iteration 9000 (5.66598 iter/s, 35.2984s/200 iter), 18.7/104ep, loss = 0.770996
  5609. I0818 14:02:49.803730 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.770996 (* 1 = 0.770996 loss)
  5610. I0818 14:02:49.803797 18523 sgd_solver.cpp:180] [0.0] Iteration 9000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5611. I0818 14:03:03.033535 19072 data_reader.cpp:321] Restarting data pre-fetching
  5612. I0818 14:03:15.173444 18523 solver.cpp:333] [0.0] Iteration 9200 (7.8838 iter/s, 25.3685s/200 iter), 19.1/104ep, loss = 1.27832
  5613. I0818 14:03:15.174124 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.27832 (* 1 = 1.27832 loss)
  5614. I0818 14:03:15.174194 18523 sgd_solver.cpp:180] [0.0] Iteration 9200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5615. I0818 14:03:40.798262 18523 solver.cpp:333] [0.0] Iteration 9400 (7.80538 iter/s, 25.6234s/200 iter), 19.6/104ep, loss = 1.33789
  5616. I0818 14:03:40.798368 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.33789 (* 1 = 1.33789 loss)
  5617. I0818 14:03:40.798382 18523 sgd_solver.cpp:180] [0.0] Iteration 9400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5618. I0818 14:04:04.846272 19072 data_reader.cpp:321] Restarting data pre-fetching
  5619. I0818 14:04:06.380671 18523 solver.cpp:333] [0.0] Iteration 9600 (7.81833 iter/s, 25.5809s/200 iter), 20/104ep, loss = 0.854004
  5620. I0818 14:04:06.380807 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.854004 (* 1 = 0.854004 loss)
  5621. I0818 14:04:06.380846 18523 sgd_solver.cpp:180] [0.0] Iteration 9600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5622. I0818 14:04:31.677048 18523 solver.cpp:333] [0.0] Iteration 9800 (7.90673 iter/s, 25.2949s/200 iter), 20.4/104ep, loss = 0.263184
  5623. I0818 14:04:31.677203 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.263184 (* 1 = 0.263184 loss)
  5624. I0818 14:04:31.677248 18523 sgd_solver.cpp:180] [0.0] Iteration 9800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5625. I0818 14:04:57.084914 18523 solver.cpp:501] Iteration 10000, Testing net (#0)
  5626. I0818 14:05:02.211519 18511 data_reader.cpp:321] Restarting data pre-fetching
  5627. I0818 14:05:06.976373 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.788398
  5628. I0818 14:05:06.976436 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.608061 (* 1 = 0.608061 loss)
  5629. I0818 14:05:06.976542 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.2975s
  5630. I0818 14:05:07.103170 18523 solver.cpp:333] [0.0] Iteration 10000 (5.66612 iter/s, 35.2975s/200 iter), 20.8/104ep, loss = 0.348145
  5631. I0818 14:05:07.103350 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.348145 (* 1 = 0.348145 loss)
  5632. I0818 14:05:07.103425 18523 sgd_solver.cpp:180] [0.0] Iteration 10000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5633. I0818 14:05:15.655553 19072 data_reader.cpp:321] Restarting data pre-fetching
  5634. I0818 14:05:32.332974 18523 solver.cpp:333] [0.0] Iteration 10200 (7.92758 iter/s, 25.2284s/200 iter), 21.2/104ep, loss = 0.442383
  5635. I0818 14:05:32.333437 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.442383 (* 1 = 0.442383 loss)
  5636. I0818 14:05:32.333465 18523 sgd_solver.cpp:180] [0.0] Iteration 10200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5637. I0818 14:05:57.386232 18523 solver.cpp:333] [0.0] Iteration 10400 (7.98346 iter/s, 25.0518s/200 iter), 21.6/104ep, loss = 0.983398
  5638. I0818 14:05:57.386358 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.983398 (* 1 = 0.983398 loss)
  5639. I0818 14:05:57.386406 18523 sgd_solver.cpp:180] [0.0] Iteration 10400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5640. I0818 14:06:15.938493 19072 data_reader.cpp:321] Restarting data pre-fetching
  5641. I0818 14:06:22.570819 18523 solver.cpp:333] [0.0] Iteration 10600 (7.9418 iter/s, 25.1832s/200 iter), 22/104ep, loss = 0.37207
  5642. I0818 14:06:22.570924 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.37207 (* 1 = 0.37207 loss)
  5643. I0818 14:06:22.570943 18523 sgd_solver.cpp:180] [0.0] Iteration 10600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5644. I0818 14:06:47.738108 18523 solver.cpp:333] [0.0] Iteration 10800 (7.94727 iter/s, 25.1659s/200 iter), 22.5/104ep, loss = 0.538086
  5645. I0818 14:06:47.738301 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.538086 (* 1 = 0.538086 loss)
  5646. I0818 14:06:47.738322 18523 sgd_solver.cpp:180] [0.0] Iteration 10800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5647. I0818 14:07:12.648247 18523 solver.cpp:501] Iteration 11000, Testing net (#0)
  5648. I0818 14:07:17.436305 18511 data_reader.cpp:321] Restarting data pre-fetching
  5649. I0818 14:07:22.023269 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.814551
  5650. I0818 14:07:22.026293 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.556495 (* 1 = 0.556495 loss)
  5651. I0818 14:07:22.026372 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.2864s
  5652. I0818 14:07:22.147389 18523 solver.cpp:333] [0.0] Iteration 11000 (5.83322 iter/s, 34.2864s/200 iter), 22.9/104ep, loss = 0.750977
  5653. I0818 14:07:22.147596 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.750977 (* 1 = 0.750977 loss)
  5654. I0818 14:07:22.147634 18523 sgd_solver.cpp:180] [0.0] Iteration 11000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5655. I0818 14:07:25.575830 19072 data_reader.cpp:321] Restarting data pre-fetching
  5656. I0818 14:07:47.019660 18523 solver.cpp:333] [0.0] Iteration 11200 (8.04154 iter/s, 24.8709s/200 iter), 23.3/104ep, loss = 0.441406
  5657. I0818 14:07:47.019842 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.441406 (* 1 = 0.441406 loss)
  5658. I0818 14:07:47.019870 18523 sgd_solver.cpp:180] [0.0] Iteration 11200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5659. I0818 14:08:12.121991 18523 solver.cpp:333] [0.0] Iteration 11400 (7.96785 iter/s, 25.1009s/200 iter), 23.7/104ep, loss = 0.184082
  5660. I0818 14:08:12.122642 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.184082 (* 1 = 0.184082 loss)
  5661. I0818 14:08:12.122686 18523 sgd_solver.cpp:180] [0.0] Iteration 11400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5662. I0818 14:08:26.106628 19072 data_reader.cpp:321] Restarting data pre-fetching
  5663. I0818 14:08:37.316678 18523 solver.cpp:333] [0.0] Iteration 11600 (7.93863 iter/s, 25.1933s/200 iter), 24.1/104ep, loss = 0.428955
  5664. I0818 14:08:37.316784 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.428955 (* 1 = 0.428955 loss)
  5665. I0818 14:08:37.316798 18523 sgd_solver.cpp:180] [0.0] Iteration 11600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5666. I0818 14:09:02.377225 18523 solver.cpp:333] [0.0] Iteration 11800 (7.98111 iter/s, 25.0592s/200 iter), 24.5/104ep, loss = 0.233154
  5667. I0818 14:09:02.377633 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.233154 (* 1 = 0.233154 loss)
  5668. I0818 14:09:02.377660 18523 sgd_solver.cpp:180] [0.0] Iteration 11800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5669. I0818 14:09:26.555289 19072 data_reader.cpp:321] Restarting data pre-fetching
  5670. I0818 14:09:27.551966 18523 solver.cpp:501] Iteration 12000, Testing net (#0)
  5671. I0818 14:09:31.961303 18511 data_reader.cpp:321] Restarting data pre-fetching
  5672. I0818 14:09:37.003746 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.795322
  5673. I0818 14:09:37.003973 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.588865 (* 1 = 0.588865 loss)
  5674. I0818 14:09:37.004045 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.6249s
  5675. I0818 14:09:37.129875 18523 solver.cpp:333] [0.0] Iteration 12000 (5.77619 iter/s, 34.6249s/200 iter), 25/104ep, loss = 0.135864
  5676. I0818 14:09:37.130220 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.135864 (* 1 = 0.135864 loss)
  5677. I0818 14:09:37.130264 18523 sgd_solver.cpp:180] [0.0] Iteration 12000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5678. I0818 14:10:02.225034 18523 solver.cpp:333] [0.0] Iteration 12200 (7.97012 iter/s, 25.0937s/200 iter), 25.4/104ep, loss = 0.0583191
  5679. I0818 14:10:02.225137 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0583191 (* 1 = 0.0583191 loss)
  5680. I0818 14:10:02.225153 18523 sgd_solver.cpp:180] [0.0] Iteration 12200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5681. I0818 14:10:27.307675 18523 solver.cpp:333] [0.0] Iteration 12400 (7.97407 iter/s, 25.0813s/200 iter), 25.8/104ep, loss = 0.580078
  5682. I0818 14:10:27.307936 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.580078 (* 1 = 0.580078 loss)
  5683. I0818 14:10:27.307952 18523 sgd_solver.cpp:180] [0.0] Iteration 12400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5684. I0818 14:10:36.284580 19072 data_reader.cpp:321] Restarting data pre-fetching
  5685. I0818 14:10:52.534284 18523 solver.cpp:333] [0.0] Iteration 12600 (7.92859 iter/s, 25.2252s/200 iter), 26.2/104ep, loss = 0.28125
  5686. I0818 14:10:52.534426 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.28125 (* 1 = 0.28125 loss)
  5687. I0818 14:10:52.534469 18523 sgd_solver.cpp:180] [0.0] Iteration 12600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5688. I0818 14:11:17.870491 18523 solver.cpp:333] [0.0] Iteration 12800 (7.89428 iter/s, 25.3348s/200 iter), 26.6/104ep, loss = 0.491943
  5689. I0818 14:11:17.871156 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.491943 (* 1 = 0.491943 loss)
  5690. I0818 14:11:17.871176 18523 sgd_solver.cpp:180] [0.0] Iteration 12800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5691. I0818 14:11:36.793578 19072 data_reader.cpp:321] Restarting data pre-fetching
  5692. I0818 14:11:42.899014 18523 solver.cpp:501] Iteration 13000, Testing net (#0)
  5693. I0818 14:11:46.962595 18511 data_reader.cpp:321] Restarting data pre-fetching
  5694. I0818 14:11:47.012117 18820 blocking_queue.cpp:40] Data layer prefetch queue empty
  5695. I0818 14:11:52.384692 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.804556
  5696. I0818 14:11:52.384878 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.599034 (* 1 = 0.599034 loss)
  5697. I0818 14:11:52.384912 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.5126s
  5698. I0818 14:11:52.507654 18523 solver.cpp:333] [0.0] Iteration 13000 (5.79499 iter/s, 34.5126s/200 iter), 27/104ep, loss = 0.661621
  5699. I0818 14:11:52.507750 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.661621 (* 1 = 0.661621 loss)
  5700. I0818 14:11:52.507776 18523 sgd_solver.cpp:180] [0.0] Iteration 13000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5701. I0818 14:12:17.571009 18523 solver.cpp:333] [0.0] Iteration 13200 (7.98024 iter/s, 25.0619s/200 iter), 27.5/104ep, loss = 0.426025
  5702. I0818 14:12:17.571252 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.426025 (* 1 = 0.426025 loss)
  5703. I0818 14:12:17.571293 18523 sgd_solver.cpp:180] [0.0] Iteration 13200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5704. I0818 14:12:42.954043 18523 solver.cpp:333] [0.0] Iteration 13400 (7.87973 iter/s, 25.3816s/200 iter), 27.9/104ep, loss = 0.202393
  5705. I0818 14:12:42.957362 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.202393 (* 1 = 0.202393 loss)
  5706. I0818 14:12:42.957387 18523 sgd_solver.cpp:180] [0.0] Iteration 13400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5707. I0818 14:12:47.422008 19072 data_reader.cpp:321] Restarting data pre-fetching
  5708. I0818 14:13:08.448036 18523 solver.cpp:333] [0.0] Iteration 13600 (7.84544 iter/s, 25.4925s/200 iter), 28.3/104ep, loss = 0.38916
  5709. I0818 14:13:08.448120 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.38916 (* 1 = 0.38916 loss)
  5710. I0818 14:13:08.448133 18523 sgd_solver.cpp:180] [0.0] Iteration 13600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5711. I0818 14:13:33.865712 18523 solver.cpp:333] [0.0] Iteration 13800 (7.86895 iter/s, 25.4164s/200 iter), 28.7/104ep, loss = 0.686035
  5712. I0818 14:13:33.868609 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.686035 (* 1 = 0.686035 loss)
  5713. I0818 14:13:33.868624 18523 sgd_solver.cpp:180] [0.0] Iteration 13800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5714. I0818 14:13:48.371364 19072 data_reader.cpp:321] Restarting data pre-fetching
  5715. I0818 14:13:59.136371 18523 solver.cpp:501] Iteration 14000, Testing net (#0)
  5716. I0818 14:14:02.913828 18511 data_reader.cpp:321] Restarting data pre-fetching
  5717. I0818 14:14:08.559582 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.773003
  5718. I0818 14:14:08.559775 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.713979 (* 1 = 0.713979 loss)
  5719. I0818 14:14:08.559856 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.6923s
  5720. I0818 14:14:08.681087 18523 solver.cpp:333] [0.0] Iteration 14000 (5.76498 iter/s, 34.6923s/200 iter), 29.1/104ep, loss = 0.238403
  5721. I0818 14:14:08.681195 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.238403 (* 1 = 0.238403 loss)
  5722. I0818 14:14:08.681241 18523 sgd_solver.cpp:180] [0.0] Iteration 14000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5723. I0818 14:14:33.690965 18523 solver.cpp:333] [0.0] Iteration 14200 (7.99731 iter/s, 25.0084s/200 iter), 29.5/104ep, loss = 0.47168
  5724. I0818 14:14:33.691059 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.47168 (* 1 = 0.47168 loss)
  5725. I0818 14:14:33.691072 18523 sgd_solver.cpp:180] [0.0] Iteration 14200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5726. I0818 14:14:58.138927 19072 data_reader.cpp:321] Restarting data pre-fetching
  5727. I0818 14:14:58.923141 18523 solver.cpp:333] [0.0] Iteration 14400 (7.92681 iter/s, 25.2308s/200 iter), 30/104ep, loss = 0.949707
  5728. I0818 14:14:58.923241 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.949707 (* 1 = 0.949707 loss)
  5729. I0818 14:14:58.923264 18523 sgd_solver.cpp:180] [0.0] Iteration 14400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5730. I0818 14:15:24.057289 18523 solver.cpp:333] [0.0] Iteration 14600 (7.95776 iter/s, 25.1327s/200 iter), 30.4/104ep, loss = 0.692383
  5731. I0818 14:15:24.057411 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.692383 (* 1 = 0.692383 loss)
  5732. I0818 14:15:24.057447 18523 sgd_solver.cpp:180] [0.0] Iteration 14600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5733. I0818 14:15:49.519688 18523 solver.cpp:333] [0.0] Iteration 14800 (7.85516 iter/s, 25.461s/200 iter), 30.8/104ep, loss = 0.208862
  5734. I0818 14:15:49.520376 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.208862 (* 1 = 0.208862 loss)
  5735. I0818 14:15:49.520406 18523 sgd_solver.cpp:180] [0.0] Iteration 14800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5736. I0818 14:15:58.820739 19072 data_reader.cpp:321] Restarting data pre-fetching
  5737. I0818 14:16:14.907830 18523 solver.cpp:501] Iteration 15000, Testing net (#0)
  5738. I0818 14:16:18.330579 18511 data_reader.cpp:321] Restarting data pre-fetching
  5739. I0818 14:16:24.411657 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.799922
  5740. I0818 14:16:24.411882 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.567185 (* 1 = 0.567185 loss)
  5741. I0818 14:16:24.411975 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.8904s
  5742. I0818 14:16:24.534518 18523 solver.cpp:333] [0.0] Iteration 15000 (5.73223 iter/s, 34.8904s/200 iter), 31.2/104ep, loss = 0.579102
  5743. I0818 14:16:24.534804 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.579102 (* 1 = 0.579102 loss)
  5744. I0818 14:16:24.534874 18523 sgd_solver.cpp:180] [0.0] Iteration 15000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5745. I0818 14:16:49.636590 18523 solver.cpp:333] [0.0] Iteration 15200 (7.96791 iter/s, 25.1007s/200 iter), 31.6/104ep, loss = 1.12109
  5746. I0818 14:16:49.636677 18523 solver.cpp:361] [0.0] Train net output #0: loss = 1.12109 (* 1 = 1.12109 loss)
  5747. I0818 14:16:49.636688 18523 sgd_solver.cpp:180] [0.0] Iteration 15200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5748. I0818 14:17:09.245949 19072 data_reader.cpp:321] Restarting data pre-fetching
  5749. I0818 14:17:14.775396 18523 solver.cpp:333] [0.0] Iteration 15400 (7.95627 iter/s, 25.1374s/200 iter), 32/104ep, loss = 0.382568
  5750. I0818 14:17:14.775519 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.382568 (* 1 = 0.382568 loss)
  5751. I0818 14:17:14.775554 18523 sgd_solver.cpp:180] [0.0] Iteration 15400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5752. I0818 14:17:40.196038 18523 solver.cpp:333] [0.0] Iteration 15600 (7.86806 iter/s, 25.4192s/200 iter), 32.4/104ep, loss = 0.522949
  5753. I0818 14:17:40.196291 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.522949 (* 1 = 0.522949 loss)
  5754. I0818 14:17:40.196319 18523 sgd_solver.cpp:180] [0.0] Iteration 15600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5755. I0818 14:18:05.403678 18523 solver.cpp:333] [0.0] Iteration 15800 (7.93454 iter/s, 25.2062s/200 iter), 32.9/104ep, loss = 0.626953
  5756. I0818 14:18:05.403846 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.626953 (* 1 = 0.626953 loss)
  5757. I0818 14:18:05.403892 18523 sgd_solver.cpp:180] [0.0] Iteration 15800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5758. I0818 14:18:10.275918 19072 data_reader.cpp:321] Restarting data pre-fetching
  5759. I0818 14:18:31.102690 18523 solver.cpp:501] Iteration 16000, Testing net (#0)
  5760. I0818 14:18:34.165235 18511 data_reader.cpp:321] Restarting data pre-fetching
  5761. I0818 14:18:40.567072 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.82147
  5762. I0818 14:18:40.567268 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.540023 (* 1 = 0.540023 loss)
  5763. I0818 14:18:40.567382 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.1618s
  5764. I0818 14:18:40.689083 18523 solver.cpp:333] [0.0] Iteration 16000 (5.68799 iter/s, 35.1618s/200 iter), 33.3/104ep, loss = 0.623047
  5765. I0818 14:18:40.689252 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.623047 (* 1 = 0.623047 loss)
  5766. I0818 14:18:40.689308 18523 sgd_solver.cpp:180] [0.0] Iteration 16000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5767. I0818 14:19:05.652177 18523 solver.cpp:333] [0.0] Iteration 16200 (8.01229 iter/s, 24.9616s/200 iter), 33.7/104ep, loss = 0.383545
  5768. I0818 14:19:05.652379 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.383545 (* 1 = 0.383545 loss)
  5769. I0818 14:19:05.652431 18523 sgd_solver.cpp:180] [0.0] Iteration 16200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5770. I0818 14:19:20.571107 19072 data_reader.cpp:321] Restarting data pre-fetching
  5771. I0818 14:19:31.116756 18523 solver.cpp:333] [0.0] Iteration 16400 (7.85447 iter/s, 25.4632s/200 iter), 34.1/104ep, loss = 0.468018
  5772. I0818 14:19:31.116871 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.468018 (* 1 = 0.468018 loss)
  5773. I0818 14:19:31.116888 18523 sgd_solver.cpp:180] [0.0] Iteration 16400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5774. I0818 14:19:56.465082 18523 solver.cpp:333] [0.0] Iteration 16600 (7.89051 iter/s, 25.3469s/200 iter), 34.5/104ep, loss = 0.0942383
  5775. I0818 14:19:56.465353 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0942383 (* 1 = 0.0942383 loss)
  5776. I0818 14:19:56.465373 18523 sgd_solver.cpp:180] [0.0] Iteration 16600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5777. I0818 14:20:21.259616 19072 data_reader.cpp:321] Restarting data pre-fetching
  5778. I0818 14:20:21.632608 18523 solver.cpp:333] [0.0] Iteration 16800 (7.94719 iter/s, 25.1661s/200 iter), 34.9/104ep, loss = 0.244873
  5779. I0818 14:20:21.632697 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.244873 (* 1 = 0.244873 loss)
  5780. I0818 14:20:21.632715 18523 sgd_solver.cpp:180] [0.0] Iteration 16800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5781. I0818 14:20:46.587868 18523 solver.cpp:501] Iteration 17000, Testing net (#0)
  5782. I0818 14:20:49.343394 18511 data_reader.cpp:321] Restarting data pre-fetching
  5783. I0818 14:20:56.495111 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.831479
  5784. I0818 14:20:56.495164 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.494965 (* 1 = 0.494965 loss)
  5785. I0818 14:20:56.495254 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.8608s
  5786. I0818 14:20:56.616127 18523 solver.cpp:333] [0.0] Iteration 17000 (5.73711 iter/s, 34.8608s/200 iter), 35.4/104ep, loss = 0.606445
  5787. I0818 14:20:56.616282 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.606445 (* 1 = 0.606445 loss)
  5788. I0818 14:20:56.616330 18523 sgd_solver.cpp:180] [0.0] Iteration 17000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5789. I0818 14:21:21.653578 18523 solver.cpp:333] [0.0] Iteration 17200 (7.98848 iter/s, 25.036s/200 iter), 35.8/104ep, loss = 0.404785
  5790. I0818 14:21:21.654042 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.404785 (* 1 = 0.404785 loss)
  5791. I0818 14:21:21.654075 18523 sgd_solver.cpp:180] [0.0] Iteration 17200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5792. I0818 14:21:31.784368 19072 data_reader.cpp:321] Restarting data pre-fetching
  5793. I0818 14:21:46.923296 18523 solver.cpp:333] [0.0] Iteration 17400 (7.91504 iter/s, 25.2684s/200 iter), 36.2/104ep, loss = 0.147827
  5794. I0818 14:21:46.923396 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.147827 (* 1 = 0.147827 loss)
  5795. I0818 14:21:46.923414 18523 sgd_solver.cpp:180] [0.0] Iteration 17400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5796. I0818 14:22:12.371685 18523 solver.cpp:333] [0.0] Iteration 17600 (7.85949 iter/s, 25.447s/200 iter), 36.6/104ep, loss = 0.129517
  5797. I0818 14:22:12.372227 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.129517 (* 1 = 0.129517 loss)
  5798. I0818 14:22:12.372319 18523 sgd_solver.cpp:180] [0.0] Iteration 17600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5799. I0818 14:22:32.727324 19072 data_reader.cpp:321] Restarting data pre-fetching
  5800. I0818 14:22:37.781813 18523 solver.cpp:333] [0.0] Iteration 17800 (7.8713 iter/s, 25.4088s/200 iter), 37/104ep, loss = 0.463135
  5801. I0818 14:22:37.781924 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.463135 (* 1 = 0.463135 loss)
  5802. I0818 14:22:37.781946 18523 sgd_solver.cpp:180] [0.0] Iteration 17800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5803. I0818 14:23:02.792641 18523 solver.cpp:501] Iteration 18000, Testing net (#0)
  5804. I0818 14:23:05.135996 18511 data_reader.cpp:321] Restarting data pre-fetching
  5805. I0818 14:23:12.380688 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.810684
  5806. I0818 14:23:12.380739 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.579411 (* 1 = 0.579411 loss)
  5807. I0818 14:23:12.380777 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.5971s
  5808. I0818 14:23:12.505239 18523 solver.cpp:333] [0.0] Iteration 18000 (5.78083 iter/s, 34.5971s/200 iter), 37.4/104ep, loss = 0.625
  5809. I0818 14:23:12.505416 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.625 (* 1 = 0.625 loss)
  5810. I0818 14:23:12.505447 18523 sgd_solver.cpp:180] [0.0] Iteration 18000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5811. I0818 14:23:37.701519 18523 solver.cpp:333] [0.0] Iteration 18200 (7.93812 iter/s, 25.1949s/200 iter), 37.9/104ep, loss = 0.238037
  5812. I0818 14:23:37.702118 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.238037 (* 1 = 0.238037 loss)
  5813. I0818 14:23:37.702174 18523 sgd_solver.cpp:180] [0.0] Iteration 18200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5814. I0818 14:23:42.701351 19072 data_reader.cpp:321] Restarting data pre-fetching
  5815. I0818 14:24:02.875887 18523 solver.cpp:333] [0.0] Iteration 18400 (7.94502 iter/s, 25.173s/200 iter), 38.3/104ep, loss = 0.998047
  5816. I0818 14:24:02.875977 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.998047 (* 1 = 0.998047 loss)
  5817. I0818 14:24:02.876027 18523 sgd_solver.cpp:180] [0.0] Iteration 18400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5818. I0818 14:24:28.230397 18523 solver.cpp:333] [0.0] Iteration 18600 (7.88857 iter/s, 25.3531s/200 iter), 38.7/104ep, loss = 0.558105
  5819. I0818 14:24:28.230834 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.558105 (* 1 = 0.558105 loss)
  5820. I0818 14:24:28.230877 18523 sgd_solver.cpp:180] [0.0] Iteration 18600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5821. I0818 14:24:43.591274 19072 data_reader.cpp:321] Restarting data pre-fetching
  5822. I0818 14:24:53.700587 18523 solver.cpp:333] [0.0] Iteration 18800 (7.8526 iter/s, 25.4693s/200 iter), 39.1/104ep, loss = 0.251465
  5823. I0818 14:24:53.700675 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.251465 (* 1 = 0.251465 loss)
  5824. I0818 14:24:53.700706 18523 sgd_solver.cpp:180] [0.0] Iteration 18800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5825. I0818 14:25:18.824185 18523 solver.cpp:501] Iteration 19000, Testing net (#0)
  5826. I0818 14:25:20.810550 18511 data_reader.cpp:321] Restarting data pre-fetching
  5827. I0818 14:25:28.361696 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.833774
  5828. I0818 14:25:28.361752 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.512961 (* 1 = 0.512961 loss)
  5829. I0818 14:25:28.361850 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.66s
  5830. I0818 14:25:28.485186 18523 solver.cpp:333] [0.0] Iteration 19000 (5.77034 iter/s, 34.66s/200 iter), 39.5/104ep, loss = 0.158081
  5831. I0818 14:25:28.485339 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.158081 (* 1 = 0.158081 loss)
  5832. I0818 14:25:28.485412 18523 sgd_solver.cpp:180] [0.0] Iteration 19000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5833. I0818 14:25:53.470955 18523 solver.cpp:333] [0.0] Iteration 19200 (8.00487 iter/s, 24.9848s/200 iter), 39.9/104ep, loss = 0.0946045
  5834. I0818 14:25:53.474028 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0946045 (* 1 = 0.0946045 loss)
  5835. I0818 14:25:53.474048 18523 sgd_solver.cpp:180] [0.0] Iteration 19200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5836. I0818 14:25:53.847638 19072 data_reader.cpp:321] Restarting data pre-fetching
  5837. I0818 14:26:18.858157 18523 solver.cpp:333] [0.0] Iteration 19400 (7.87829 iter/s, 25.3862s/200 iter), 40.4/104ep, loss = 0.230835
  5838. I0818 14:26:18.858243 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.230835 (* 1 = 0.230835 loss)
  5839. I0818 14:26:18.858255 18523 sgd_solver.cpp:180] [0.0] Iteration 19400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5840. I0818 14:26:44.117511 18523 solver.cpp:333] [0.0] Iteration 19600 (7.91816 iter/s, 25.2584s/200 iter), 40.8/104ep, loss = 0.567383
  5841. I0818 14:26:44.119340 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.567383 (* 1 = 0.567383 loss)
  5842. I0818 14:26:44.119359 18523 sgd_solver.cpp:180] [0.0] Iteration 19600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5843. I0818 14:26:54.719157 19072 data_reader.cpp:321] Restarting data pre-fetching
  5844. I0818 14:27:09.587277 18523 solver.cpp:333] [0.0] Iteration 19800 (7.85278 iter/s, 25.4687s/200 iter), 41.2/104ep, loss = 0.0563965
  5845. I0818 14:27:09.587450 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0563965 (* 1 = 0.0563965 loss)
  5846. I0818 14:27:09.587476 18523 sgd_solver.cpp:180] [0.0] Iteration 19800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5847. I0818 14:27:34.886164 18523 solver.cpp:501] Iteration 20000, Testing net (#0)
  5848. I0818 14:27:36.607470 18511 data_reader.cpp:321] Restarting data pre-fetching
  5849. I0818 14:27:44.351406 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.804551
  5850. I0818 14:27:44.351475 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.596144 (* 1 = 0.596144 loss)
  5851. I0818 14:27:44.351631 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.7629s
  5852. I0818 14:27:44.473821 18523 solver.cpp:333] [0.0] Iteration 20000 (5.75326 iter/s, 34.7629s/200 iter), 41.6/104ep, loss = 0.479736
  5853. I0818 14:27:44.473942 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.479736 (* 1 = 0.479736 loss)
  5854. I0818 14:27:44.473999 18523 sgd_solver.cpp:180] [0.0] Iteration 20000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5855. I0818 14:28:04.925303 19072 data_reader.cpp:321] Restarting data pre-fetching
  5856. I0818 14:28:09.528077 18523 solver.cpp:333] [0.0] Iteration 20200 (7.98303 iter/s, 25.0532s/200 iter), 42/104ep, loss = 0.852051
  5857. I0818 14:28:09.528275 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.852051 (* 1 = 0.852051 loss)
  5858. I0818 14:28:09.528321 18523 sgd_solver.cpp:180] [0.0] Iteration 20200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5859. I0818 14:28:34.727861 18523 solver.cpp:333] [0.0] Iteration 20400 (7.9369 iter/s, 25.1987s/200 iter), 42.4/104ep, loss = 0.0669556
  5860. I0818 14:28:34.727969 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0669556 (* 1 = 0.0669556 loss)
  5861. I0818 14:28:34.728025 18523 sgd_solver.cpp:180] [0.0] Iteration 20400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5862. I0818 14:29:00.009136 18523 solver.cpp:333] [0.0] Iteration 20600 (7.91134 iter/s, 25.2802s/200 iter), 42.8/104ep, loss = 0.459717
  5863. I0818 14:29:00.009874 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.459717 (* 1 = 0.459717 loss)
  5864. I0818 14:29:00.009905 18523 sgd_solver.cpp:180] [0.0] Iteration 20600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5865. I0818 14:29:05.867394 19072 data_reader.cpp:321] Restarting data pre-fetching
  5866. I0818 14:29:25.951102 18523 solver.cpp:333] [0.0] Iteration 20800 (7.70984 iter/s, 25.9409s/200 iter), 43.3/104ep, loss = 0.00894928
  5867. I0818 14:29:25.951208 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00894928 (* 1 = 0.00894928 loss)
  5868. I0818 14:29:25.951225 18523 sgd_solver.cpp:180] [0.0] Iteration 20800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5869. I0818 14:29:51.627517 18523 solver.cpp:501] Iteration 21000, Testing net (#0)
  5870. I0818 14:29:52.989430 18511 data_reader.cpp:321] Restarting data pre-fetching
  5871. I0818 14:30:00.972460 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.820684
  5872. I0818 14:30:00.972502 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.537973 (* 1 = 0.537973 loss)
  5873. I0818 14:30:00.972584 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.02s
  5874. I0818 14:30:01.095854 18523 solver.cpp:333] [0.0] Iteration 21000 (5.71103 iter/s, 35.02s/200 iter), 43.7/104ep, loss = 0.541504
  5875. I0818 14:30:01.096047 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.541504 (* 1 = 0.541504 loss)
  5876. I0818 14:30:01.096084 18523 sgd_solver.cpp:180] [0.0] Iteration 21000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5877. I0818 14:30:16.906087 19072 data_reader.cpp:321] Restarting data pre-fetching
  5878. I0818 14:30:26.443307 18523 solver.cpp:333] [0.0] Iteration 21200 (7.89069 iter/s, 25.3463s/200 iter), 44.1/104ep, loss = 0.0646362
  5879. I0818 14:30:26.446193 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0646362 (* 1 = 0.0646362 loss)
  5880. I0818 14:30:26.446211 18523 sgd_solver.cpp:180] [0.0] Iteration 21200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5881. I0818 14:30:52.055022 18523 solver.cpp:333] [0.0] Iteration 21400 (7.80928 iter/s, 25.6106s/200 iter), 44.5/104ep, loss = 0.219727
  5882. I0818 14:30:52.055115 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.219727 (* 1 = 0.219727 loss)
  5883. I0818 14:30:52.055133 18523 sgd_solver.cpp:180] [0.0] Iteration 21400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5884. I0818 14:31:17.745039 18523 solver.cpp:333] [0.0] Iteration 21600 (7.78549 iter/s, 25.6888s/200 iter), 44.9/104ep, loss = 0.0161285
  5885. I0818 14:31:17.745296 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0161285 (* 1 = 0.0161285 loss)
  5886. I0818 14:31:17.745342 18523 sgd_solver.cpp:180] [0.0] Iteration 21600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5887. I0818 14:31:18.507513 19072 data_reader.cpp:321] Restarting data pre-fetching
  5888. I0818 14:31:43.423280 18523 solver.cpp:333] [0.0] Iteration 21800 (7.78903 iter/s, 25.6771s/200 iter), 45.3/104ep, loss = 0.914062
  5889. I0818 14:31:43.423466 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.914062 (* 1 = 0.914062 loss)
  5890. I0818 14:31:43.423511 18523 sgd_solver.cpp:180] [0.0] Iteration 21800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5891. I0818 14:32:08.904549 18523 solver.cpp:501] Iteration 22000, Testing net (#0)
  5892. I0818 14:32:10.005513 18511 data_reader.cpp:321] Restarting data pre-fetching
  5893. I0818 14:32:18.693739 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.820679
  5894. I0818 14:32:18.693800 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.561812 (* 1 = 0.561812 loss)
  5895. I0818 14:32:18.693827 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.269s
  5896. I0818 14:32:18.817976 18523 solver.cpp:333] [0.0] Iteration 22000 (5.6707 iter/s, 35.269s/200 iter), 45.8/104ep, loss = 0.164917
  5897. I0818 14:32:18.818140 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.164917 (* 1 = 0.164917 loss)
  5898. I0818 14:32:18.818219 18523 sgd_solver.cpp:180] [0.0] Iteration 22000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5899. I0818 14:32:29.526403 19072 data_reader.cpp:321] Restarting data pre-fetching
  5900. I0818 14:32:44.176440 18523 solver.cpp:333] [0.0] Iteration 22200 (7.88728 iter/s, 25.3573s/200 iter), 46.2/104ep, loss = 0.231079
  5901. I0818 14:32:44.176692 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.231079 (* 1 = 0.231079 loss)
  5902. I0818 14:32:44.176740 18523 sgd_solver.cpp:180] [0.0] Iteration 22200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5903. I0818 14:33:09.806597 18523 solver.cpp:333] [0.0] Iteration 22400 (7.80367 iter/s, 25.629s/200 iter), 46.6/104ep, loss = 0.09198
  5904. I0818 14:33:09.806777 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.09198 (* 1 = 0.09198 loss)
  5905. I0818 14:33:09.806828 18523 sgd_solver.cpp:180] [0.0] Iteration 22400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5906. I0818 14:33:31.512511 19072 data_reader.cpp:321] Restarting data pre-fetching
  5907. I0818 14:33:35.428898 18523 solver.cpp:333] [0.0] Iteration 22600 (7.80606 iter/s, 25.6211s/200 iter), 47/104ep, loss = 0.0270233
  5908. I0818 14:33:35.429014 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0270233 (* 1 = 0.0270233 loss)
  5909. I0818 14:33:35.429038 18523 sgd_solver.cpp:180] [0.0] Iteration 22600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5910. I0818 14:34:00.609321 18523 solver.cpp:333] [0.0] Iteration 22800 (7.94361 iter/s, 25.1775s/200 iter), 47.4/104ep, loss = 0.422119
  5911. I0818 14:34:00.609448 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.422119 (* 1 = 0.422119 loss)
  5912. I0818 14:34:00.609469 18523 sgd_solver.cpp:180] [0.0] Iteration 22800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5913. I0818 14:34:26.520758 18523 solver.cpp:501] Iteration 23000, Testing net (#0)
  5914. I0818 14:34:27.170835 18511 data_reader.cpp:321] Restarting data pre-fetching
  5915. I0818 14:34:36.176105 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.816851
  5916. I0818 14:34:36.176142 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.522904 (* 1 = 0.522904 loss)
  5917. I0818 14:34:36.176239 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.5671s
  5918. I0818 14:34:36.298099 18523 solver.cpp:333] [0.0] Iteration 23000 (5.62318 iter/s, 35.5671s/200 iter), 47.8/104ep, loss = 0.445557
  5919. I0818 14:34:36.298245 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.445557 (* 1 = 0.445557 loss)
  5920. I0818 14:34:36.298265 18523 sgd_solver.cpp:180] [0.0] Iteration 23000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5921. I0818 14:34:42.530517 19072 data_reader.cpp:321] Restarting data pre-fetching
  5922. I0818 14:35:01.510263 18523 solver.cpp:333] [0.0] Iteration 23200 (7.93307 iter/s, 25.2109s/200 iter), 48.3/104ep, loss = 0.24231
  5923. I0818 14:35:01.510514 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.24231 (* 1 = 0.24231 loss)
  5924. I0818 14:35:01.510535 18523 sgd_solver.cpp:180] [0.0] Iteration 23200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5925. I0818 14:35:27.001665 18523 solver.cpp:333] [0.0] Iteration 23400 (7.84616 iter/s, 25.4902s/200 iter), 48.7/104ep, loss = 0.0872803
  5926. I0818 14:35:27.001972 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0872803 (* 1 = 0.0872803 loss)
  5927. I0818 14:35:27.002020 18523 sgd_solver.cpp:180] [0.0] Iteration 23400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5928. I0818 14:35:43.448882 19072 data_reader.cpp:321] Restarting data pre-fetching
  5929. I0818 14:35:52.519340 18523 solver.cpp:333] [0.0] Iteration 23600 (7.83809 iter/s, 25.5164s/200 iter), 49.1/104ep, loss = 0.249023
  5930. I0818 14:35:52.519512 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.249023 (* 1 = 0.249023 loss)
  5931. I0818 14:35:52.519552 18523 sgd_solver.cpp:180] [0.0] Iteration 23600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5932. I0818 14:36:18.264351 18523 solver.cpp:333] [0.0] Iteration 23800 (7.76887 iter/s, 25.7438s/200 iter), 49.5/104ep, loss = 0.623535
  5933. I0818 14:36:18.264577 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.623535 (* 1 = 0.623535 loss)
  5934. I0818 14:36:18.264619 18523 sgd_solver.cpp:180] [0.0] Iteration 23800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5935. I0818 14:36:43.727425 18523 solver.cpp:501] Iteration 24000, Testing net (#0)
  5936. I0818 14:36:44.001840 18511 data_reader.cpp:321] Restarting data pre-fetching
  5937. I0818 14:36:52.595274 18511 data_reader.cpp:321] Restarting data pre-fetching
  5938. I0818 14:36:53.348145 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.774551
  5939. I0818 14:36:53.348197 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.747698 (* 1 = 0.747698 loss)
  5940. I0818 14:36:53.348263 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.0823s
  5941. I0818 14:36:53.477988 18523 solver.cpp:333] [0.0] Iteration 24000 (5.70089 iter/s, 35.0823s/200 iter), 49.9/104ep, loss = 0.370361
  5942. I0818 14:36:53.478181 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.370361 (* 1 = 0.370361 loss)
  5943. I0818 14:36:53.478247 18523 sgd_solver.cpp:180] [0.0] Iteration 24000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5944. I0818 14:36:54.614481 19072 data_reader.cpp:321] Restarting data pre-fetching
  5945. I0818 14:37:18.721813 18523 solver.cpp:333] [0.0] Iteration 24200 (7.92312 iter/s, 25.2426s/200 iter), 50.3/104ep, loss = 0.417236
  5946. I0818 14:37:18.721935 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.417236 (* 1 = 0.417236 loss)
  5947. I0818 14:37:18.721952 18523 sgd_solver.cpp:180] [0.0] Iteration 24200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5948. I0818 14:37:44.086591 18523 solver.cpp:333] [0.0] Iteration 24400 (7.88536 iter/s, 25.3635s/200 iter), 50.8/104ep, loss = 0.549805
  5949. I0818 14:37:44.086838 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.549805 (* 1 = 0.549805 loss)
  5950. I0818 14:37:44.086860 18523 sgd_solver.cpp:180] [0.0] Iteration 24400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5951. I0818 14:37:55.804159 19072 data_reader.cpp:321] Restarting data pre-fetching
  5952. I0818 14:38:09.682790 18523 solver.cpp:333] [0.0] Iteration 24600 (7.81404 iter/s, 25.595s/200 iter), 51.2/104ep, loss = 0.0473022
  5953. I0818 14:38:09.682915 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0473022 (* 1 = 0.0473022 loss)
  5954. I0818 14:38:09.682950 18523 sgd_solver.cpp:180] [0.0] Iteration 24600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5955. I0818 14:38:35.445590 18523 solver.cpp:333] [0.0] Iteration 24800 (7.7635 iter/s, 25.7616s/200 iter), 51.6/104ep, loss = 0.0622864
  5956. I0818 14:38:35.448164 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0622864 (* 1 = 0.0622864 loss)
  5957. I0818 14:38:35.448200 18523 sgd_solver.cpp:180] [0.0] Iteration 24800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5958. I0818 14:38:57.483139 19072 data_reader.cpp:321] Restarting data pre-fetching
  5959. I0818 14:39:00.989725 18523 solver.cpp:501] Iteration 25000, Testing net (#0)
  5960. I0818 14:39:09.516409 18511 data_reader.cpp:321] Restarting data pre-fetching
  5961. I0818 14:39:09.619940 18740 blocking_queue.cpp:40] Data layer prefetch queue empty
  5962. I0818 14:39:10.255457 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.866851
  5963. I0818 14:39:10.255508 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.418971 (* 1 = 0.418971 loss)
  5964. I0818 14:39:10.255578 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.8083s
  5965. I0818 14:39:10.379071 18523 solver.cpp:333] [0.0] Iteration 25000 (5.74575 iter/s, 34.8083s/200 iter), 52/104ep, loss = 0.319336
  5966. I0818 14:39:10.379153 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.319336 (* 1 = 0.319336 loss)
  5967. I0818 14:39:10.379169 18523 sgd_solver.cpp:180] [0.0] Iteration 25000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5968. I0818 14:39:35.515332 18523 solver.cpp:333] [0.0] Iteration 25200 (7.95703 iter/s, 25.135s/200 iter), 52.4/104ep, loss = 0.119019
  5969. I0818 14:39:35.515522 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.119019 (* 1 = 0.119019 loss)
  5970. I0818 14:39:35.515539 18523 sgd_solver.cpp:180] [0.0] Iteration 25200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5971. I0818 14:40:01.372802 18523 solver.cpp:333] [0.0] Iteration 25400 (7.7351 iter/s, 25.8562s/200 iter), 52.8/104ep, loss = 0.168579
  5972. I0818 14:40:01.373191 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.168579 (* 1 = 0.168579 loss)
  5973. I0818 14:40:01.373221 18523 sgd_solver.cpp:180] [0.0] Iteration 25400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5974. I0818 14:40:08.036382 19072 data_reader.cpp:321] Restarting data pre-fetching
  5975. I0818 14:40:26.985226 18523 solver.cpp:333] [0.0] Iteration 25600 (7.80909 iter/s, 25.6112s/200 iter), 53.2/104ep, loss = 0.212769
  5976. I0818 14:40:26.985373 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.212769 (* 1 = 0.212769 loss)
  5977. I0818 14:40:26.985405 18523 sgd_solver.cpp:180] [0.0] Iteration 25600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5978. I0818 14:40:52.296543 18523 solver.cpp:333] [0.0] Iteration 25800 (7.902 iter/s, 25.3101s/200 iter), 53.7/104ep, loss = 0.32373
  5979. I0818 14:40:52.296921 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.32373 (* 1 = 0.32373 loss)
  5980. I0818 14:40:52.296950 18523 sgd_solver.cpp:180] [0.0] Iteration 25800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5981. I0818 14:41:09.032007 19072 data_reader.cpp:321] Restarting data pre-fetching
  5982. I0818 14:41:17.595048 18523 solver.cpp:501] Iteration 26000, Testing net (#0)
  5983. I0818 14:41:25.832548 18511 data_reader.cpp:321] Restarting data pre-fetching
  5984. I0818 14:41:27.547667 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.778396
  5985. I0818 14:41:27.547726 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.808081 (* 1 = 0.808081 loss)
  5986. I0818 14:41:27.547817 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.2496s
  5987. I0818 14:41:27.675135 18523 solver.cpp:333] [0.0] Iteration 26000 (5.67383 iter/s, 35.2496s/200 iter), 54.1/104ep, loss = 0.0880127
  5988. I0818 14:41:27.675420 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0880127 (* 1 = 0.0880127 loss)
  5989. I0818 14:41:27.675463 18523 sgd_solver.cpp:180] [0.0] Iteration 26000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5990. I0818 14:41:53.206594 18523 solver.cpp:333] [0.0] Iteration 26200 (7.83388 iter/s, 25.5301s/200 iter), 54.5/104ep, loss = 0.155029
  5991. I0818 14:41:53.206699 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.155029 (* 1 = 0.155029 loss)
  5992. I0818 14:41:53.206715 18523 sgd_solver.cpp:180] [0.0] Iteration 26200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5993. I0818 14:42:18.835896 18523 solver.cpp:333] [0.0] Iteration 26400 (7.80394 iter/s, 25.6281s/200 iter), 54.9/104ep, loss = 0.0786133
  5994. I0818 14:42:18.836143 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0786133 (* 1 = 0.0786133 loss)
  5995. I0818 14:42:18.836179 18523 sgd_solver.cpp:180] [0.0] Iteration 26400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  5996. I0818 14:42:20.769973 19072 data_reader.cpp:321] Restarting data pre-fetching
  5997. I0818 14:42:44.314666 18523 solver.cpp:333] [0.0] Iteration 26600 (7.85008 iter/s, 25.4774s/200 iter), 55.3/104ep, loss = 0.0831299
  5998. I0818 14:42:44.314790 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0831299 (* 1 = 0.0831299 loss)
  5999. I0818 14:42:44.314841 18523 sgd_solver.cpp:180] [0.0] Iteration 26600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6000. I0818 14:43:09.852975 18523 solver.cpp:333] [0.0] Iteration 26800 (7.83177 iter/s, 25.537s/200 iter), 55.7/104ep, loss = 0.106567
  6001. I0818 14:43:09.853881 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.106567 (* 1 = 0.106567 loss)
  6002. I0818 14:43:09.853919 18523 sgd_solver.cpp:180] [0.0] Iteration 26800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6003. I0818 14:43:21.996718 19072 data_reader.cpp:321] Restarting data pre-fetching
  6004. I0818 14:43:35.274715 18523 solver.cpp:501] Iteration 27000, Testing net (#0)
  6005. I0818 14:43:43.180666 18511 data_reader.cpp:321] Restarting data pre-fetching
  6006. I0818 14:43:44.846796 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.840698
  6007. I0818 14:43:44.846850 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.48604 (* 1 = 0.48604 loss)
  6008. I0818 14:43:44.846958 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.9923s
  6009. I0818 14:43:44.972522 18523 solver.cpp:333] [0.0] Iteration 27000 (5.71555 iter/s, 34.9923s/200 iter), 56.2/104ep, loss = 0.319824
  6010. I0818 14:43:44.972649 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.319824 (* 1 = 0.319824 loss)
  6011. I0818 14:43:44.972676 18523 sgd_solver.cpp:180] [0.0] Iteration 27000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6012. I0818 14:44:10.121790 18523 solver.cpp:333] [0.0] Iteration 27200 (7.95294 iter/s, 25.1479s/200 iter), 56.6/104ep, loss = 0.495361
  6013. I0818 14:44:10.121989 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.495361 (* 1 = 0.495361 loss)
  6014. I0818 14:44:10.122334 18523 sgd_solver.cpp:180] [0.0] Iteration 27200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6015. I0818 14:44:32.538633 19072 data_reader.cpp:321] Restarting data pre-fetching
  6016. I0818 14:44:35.809796 18523 solver.cpp:333] [0.0] Iteration 27400 (7.78611 iter/s, 25.6868s/200 iter), 57/104ep, loss = 0.424072
  6017. I0818 14:44:35.809893 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.424072 (* 1 = 0.424072 loss)
  6018. I0818 14:44:35.809911 18523 sgd_solver.cpp:180] [0.0] Iteration 27400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6019. I0818 14:45:01.290527 18523 solver.cpp:333] [0.0] Iteration 27600 (7.84947 iter/s, 25.4794s/200 iter), 57.4/104ep, loss = 0.132935
  6020. I0818 14:45:01.290696 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.132935 (* 1 = 0.132935 loss)
  6021. I0818 14:45:01.290751 18523 sgd_solver.cpp:180] [0.0] Iteration 27600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6022. I0818 14:45:27.096663 18523 solver.cpp:333] [0.0] Iteration 27800 (7.75049 iter/s, 25.8048s/200 iter), 57.8/104ep, loss = 0.182739
  6023. I0818 14:45:27.098052 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.182739 (* 1 = 0.182739 loss)
  6024. I0818 14:45:27.098094 18523 sgd_solver.cpp:180] [0.0] Iteration 27800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6025. I0818 14:45:33.980994 19072 data_reader.cpp:321] Restarting data pre-fetching
  6026. I0818 14:45:52.407865 18523 solver.cpp:501] Iteration 28000, Testing net (#0)
  6027. I0818 14:45:59.949404 18511 data_reader.cpp:321] Restarting data pre-fetching
  6028. I0818 14:46:02.257297 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.815313
  6029. I0818 14:46:02.257351 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.554128 (* 1 = 0.554128 loss)
  6030. I0818 14:46:02.257462 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.1591s
  6031. I0818 14:46:02.382848 18523 solver.cpp:333] [0.0] Iteration 28000 (5.68843 iter/s, 35.1591s/200 iter), 58.2/104ep, loss = 0.0159454
  6032. I0818 14:46:02.382963 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0159454 (* 1 = 0.0159454 loss)
  6033. I0818 14:46:02.382985 18523 sgd_solver.cpp:180] [0.0] Iteration 28000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6034. I0818 14:46:27.780661 18523 solver.cpp:333] [0.0] Iteration 28200 (7.87511 iter/s, 25.3965s/200 iter), 58.7/104ep, loss = 0.15332
  6035. I0818 14:46:27.780824 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.15332 (* 1 = 0.15332 loss)
  6036. I0818 14:46:27.780887 18523 sgd_solver.cpp:180] [0.0] Iteration 28200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6037. I0818 14:46:45.504443 19072 data_reader.cpp:321] Restarting data pre-fetching
  6038. I0818 14:46:53.455642 18523 solver.cpp:333] [0.0] Iteration 28400 (7.79007 iter/s, 25.6737s/200 iter), 59.1/104ep, loss = 0.0211182
  6039. I0818 14:46:53.455838 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0211182 (* 1 = 0.0211182 loss)
  6040. I0818 14:46:53.455876 18523 sgd_solver.cpp:180] [0.0] Iteration 28400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6041. I0818 14:47:19.090104 18523 solver.cpp:333] [0.0] Iteration 28600 (7.80239 iter/s, 25.6332s/200 iter), 59.5/104ep, loss = 0.408447
  6042. I0818 14:47:19.091168 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.408447 (* 1 = 0.408447 loss)
  6043. I0818 14:47:19.091202 18523 sgd_solver.cpp:180] [0.0] Iteration 28600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6044. I0818 14:47:44.558214 18523 solver.cpp:333] [0.0] Iteration 28800 (7.85337 iter/s, 25.4668s/200 iter), 59.9/104ep, loss = 0.0406799
  6045. I0818 14:47:44.558331 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0406799 (* 1 = 0.0406799 loss)
  6046. I0818 14:47:44.558367 18523 sgd_solver.cpp:180] [0.0] Iteration 28800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6047. I0818 14:47:46.829581 19072 data_reader.cpp:321] Restarting data pre-fetching
  6048. I0818 14:48:09.913178 18523 solver.cpp:501] Iteration 29000, Testing net (#0)
  6049. I0818 14:48:17.105409 18511 data_reader.cpp:321] Restarting data pre-fetching
  6050. I0818 14:48:19.568863 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.813003
  6051. I0818 14:48:19.568919 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.604822 (* 1 = 0.604822 loss)
  6052. I0818 14:48:19.569010 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.0091s
  6053. I0818 14:48:19.693459 18523 solver.cpp:333] [0.0] Iteration 29000 (5.7128 iter/s, 35.0091s/200 iter), 60.3/104ep, loss = 0.293701
  6054. I0818 14:48:19.693552 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.293701 (* 1 = 0.293701 loss)
  6055. I0818 14:48:19.693598 18523 sgd_solver.cpp:180] [0.0] Iteration 29000, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6056. I0818 14:48:44.905055 18523 solver.cpp:333] [0.0] Iteration 29200 (7.93327 iter/s, 25.2103s/200 iter), 60.7/104ep, loss = 0.617188
  6057. I0818 14:48:44.905503 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.617188 (* 1 = 0.617188 loss)
  6058. I0818 14:48:44.905530 18523 sgd_solver.cpp:180] [0.0] Iteration 29200, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6059. I0818 14:48:57.176779 19072 data_reader.cpp:321] Restarting data pre-fetching
  6060. I0818 14:49:10.068174 18523 solver.cpp:333] [0.0] Iteration 29400 (7.94855 iter/s, 25.1618s/200 iter), 61.2/104ep, loss = 0.316406
  6061. I0818 14:49:10.068279 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.316406 (* 1 = 0.316406 loss)
  6062. I0818 14:49:10.068307 18523 sgd_solver.cpp:180] [0.0] Iteration 29400, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6063. I0818 14:49:35.289757 18523 solver.cpp:333] [0.0] Iteration 29600 (7.93013 iter/s, 25.2203s/200 iter), 61.6/104ep, loss = 0.357666
  6064. I0818 14:49:35.289991 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.357666 (* 1 = 0.357666 loss)
  6065. I0818 14:49:35.290006 18523 sgd_solver.cpp:180] [0.0] Iteration 29600, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6066. I0818 14:49:57.638797 19072 data_reader.cpp:321] Restarting data pre-fetching
  6067. I0818 14:50:00.424831 18523 solver.cpp:333] [0.0] Iteration 29800 (7.95742 iter/s, 25.1338s/200 iter), 62/104ep, loss = 0.0220642
  6068. I0818 14:50:00.424966 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0220642 (* 1 = 0.0220642 loss)
  6069. I0818 14:50:00.425040 18523 sgd_solver.cpp:180] [0.0] Iteration 29800, lr = 0.001, m = 0.9, lrm = 0.01, wd = 0.004, gs = 2000
  6070. I0818 14:50:25.661954 18523 solver.cpp:501] Iteration 30000, Testing net (#0)
  6071. I0818 14:50:32.500609 18511 data_reader.cpp:321] Restarting data pre-fetching
  6072. I0818 14:50:35.086452 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.852227
  6073. I0818 14:50:35.086489 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.462228 (* 1 = 0.462228 loss)
  6074. I0818 14:50:35.086588 18523 solver.cpp:271] [MultiGPU] Tests completed in 34.6601s
  6075. I0818 14:50:35.092326 19005 sgd_solver.cpp:50] MultiStep Status: Iteration 30000, step = 1
  6076. I0818 14:50:35.092631 19009 sgd_solver.cpp:50] MultiStep Status: Iteration 30000, step = 1
  6077. I0818 14:50:35.093008 19011 sgd_solver.cpp:50] MultiStep Status: Iteration 30000, step = 1
  6078. I0818 14:50:35.093053 19006 sgd_solver.cpp:50] MultiStep Status: Iteration 30000, step = 1
  6079. I0818 14:50:35.093106 19004 sgd_solver.cpp:50] MultiStep Status: Iteration 30000, step = 1
  6080. I0818 14:50:35.093179 19007 sgd_solver.cpp:50] MultiStep Status: Iteration 30000, step = 1
  6081. I0818 14:50:35.093374 19008 sgd_solver.cpp:50] MultiStep Status: Iteration 30000, step = 1
  6082. I0818 14:50:35.093538 19010 sgd_solver.cpp:50] MultiStep Status: Iteration 30000, step = 1
  6083. I0818 14:50:35.209651 18523 solver.cpp:333] [0.0] Iteration 30000 (5.77033 iter/s, 34.6601s/200 iter), 62.4/104ep, loss = 0.0853271
  6084. I0818 14:50:35.209771 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0853271 (* 1 = 0.0853271 loss)
  6085. I0818 14:50:35.209798 18523 sgd_solver.cpp:180] [0.0] Iteration 30000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6086. I0818 14:51:01.595669 18523 solver.cpp:333] [0.0] Iteration 30200 (7.58017 iter/s, 26.3846s/200 iter), 62.8/104ep, loss = 0.44458
  6087. I0818 14:51:01.596159 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.44458 (* 1 = 0.44458 loss)
  6088. I0818 14:51:01.596175 18523 sgd_solver.cpp:180] [0.0] Iteration 30200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6089. I0818 14:51:09.792574 19072 data_reader.cpp:321] Restarting data pre-fetching
  6090. I0818 14:51:28.563472 18523 solver.cpp:333] [0.0] Iteration 30400 (7.41664 iter/s, 26.9664s/200 iter), 63.2/104ep, loss = 0.0575562
  6091. I0818 14:51:28.563585 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0575562 (* 1 = 0.0575562 loss)
  6092. I0818 14:51:28.563618 18523 sgd_solver.cpp:180] [0.0] Iteration 30400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6093. I0818 14:51:55.348037 18523 solver.cpp:333] [0.0] Iteration 30600 (7.46737 iter/s, 26.7832s/200 iter), 63.6/104ep, loss = 0.445801
  6094. I0818 14:51:55.348556 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.445801 (* 1 = 0.445801 loss)
  6095. I0818 14:51:55.348608 18523 sgd_solver.cpp:180] [0.0] Iteration 30600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6096. I0818 14:52:14.558178 19072 data_reader.cpp:321] Restarting data pre-fetching
  6097. I0818 14:52:22.679653 18523 solver.cpp:333] [0.0] Iteration 30800 (7.31791 iter/s, 27.3302s/200 iter), 64.1/104ep, loss = 0.0395813
  6098. I0818 14:52:22.679805 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0395813 (* 1 = 0.0395813 loss)
  6099. I0818 14:52:22.679836 18523 sgd_solver.cpp:180] [0.0] Iteration 30800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6100. I0818 14:52:49.486915 18523 solver.cpp:501] Iteration 31000, Testing net (#0)
  6101. I0818 14:52:55.963634 18511 data_reader.cpp:321] Restarting data pre-fetching
  6102. I0818 14:52:59.120716 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.884531
  6103. I0818 14:52:59.120776 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.344724 (* 1 = 0.344724 loss)
  6104. I0818 14:52:59.120887 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.4394s
  6105. I0818 14:52:59.251087 18523 solver.cpp:333] [0.0] Iteration 31000 (5.48856 iter/s, 36.4394s/200 iter), 64.5/104ep, loss = 0.301025
  6106. I0818 14:52:59.251205 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.301025 (* 1 = 0.301025 loss)
  6107. I0818 14:52:59.251264 18523 sgd_solver.cpp:180] [0.0] Iteration 31000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6108. I0818 14:53:26.118556 18523 solver.cpp:333] [0.0] Iteration 31200 (7.44433 iter/s, 26.8661s/200 iter), 64.9/104ep, loss = 0.00937653
  6109. I0818 14:53:26.119159 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00937653 (* 1 = 0.00937653 loss)
  6110. I0818 14:53:26.119240 18523 sgd_solver.cpp:180] [0.0] Iteration 31200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6111. I0818 14:53:28.961621 19072 data_reader.cpp:321] Restarting data pre-fetching
  6112. I0818 14:53:53.122572 18523 solver.cpp:333] [0.0] Iteration 31400 (7.40668 iter/s, 27.0026s/200 iter), 65.3/104ep, loss = 0.0583191
  6113. I0818 14:53:53.122717 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0583191 (* 1 = 0.0583191 loss)
  6114. I0818 14:53:53.122751 18523 sgd_solver.cpp:180] [0.0] Iteration 31400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6115. I0818 14:54:20.047377 18523 solver.cpp:333] [0.0] Iteration 31600 (7.42847 iter/s, 26.9234s/200 iter), 65.7/104ep, loss = 0.00270462
  6116. I0818 14:54:20.047677 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00270462 (* 1 = 0.00270462 loss)
  6117. I0818 14:54:20.047705 18523 sgd_solver.cpp:180] [0.0] Iteration 31600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6118. I0818 14:54:33.590445 19072 data_reader.cpp:321] Restarting data pre-fetching
  6119. I0818 14:54:47.135321 18523 solver.cpp:333] [0.0] Iteration 31800 (7.38374 iter/s, 27.0865s/200 iter), 66.1/104ep, loss = 0.0546875
  6120. I0818 14:54:47.135498 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0546875 (* 1 = 0.0546875 loss)
  6121. I0818 14:54:47.135548 18523 sgd_solver.cpp:180] [0.0] Iteration 31800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6122. I0818 14:55:14.413601 18523 solver.cpp:501] Iteration 32000, Testing net (#0)
  6123. I0818 14:55:20.611001 18511 data_reader.cpp:321] Restarting data pre-fetching
  6124. I0818 14:55:23.842017 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.90144
  6125. I0818 14:55:23.842061 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.300955 (* 1 = 0.300955 loss)
  6126. I0818 14:55:23.842146 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.705s
  6127. I0818 14:55:23.973379 18523 solver.cpp:333] [0.0] Iteration 32000 (5.44885 iter/s, 36.705s/200 iter), 66.6/104ep, loss = 0.00355339
  6128. I0818 14:55:23.973532 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00355339 (* 1 = 0.00355339 loss)
  6129. I0818 14:55:23.973578 18523 sgd_solver.cpp:180] [0.0] Iteration 32000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6130. I0818 14:55:49.062300 19072 data_reader.cpp:321] Restarting data pre-fetching
  6131. I0818 14:55:51.260133 18523 solver.cpp:333] [0.0] Iteration 32200 (7.32995 iter/s, 27.2853s/200 iter), 67/104ep, loss = 0.11969
  6132. I0818 14:55:51.260268 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.11969 (* 1 = 0.11969 loss)
  6133. I0818 14:55:51.260303 18523 sgd_solver.cpp:180] [0.0] Iteration 32200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6134. I0818 14:56:18.417758 18523 solver.cpp:333] [0.0] Iteration 32400 (7.3648 iter/s, 27.1562s/200 iter), 67.4/104ep, loss = 0.0137177
  6135. I0818 14:56:18.417856 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0137177 (* 1 = 0.0137177 loss)
  6136. I0818 14:56:18.417871 18523 sgd_solver.cpp:180] [0.0] Iteration 32400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6137. I0818 14:56:45.654228 18523 solver.cpp:333] [0.0] Iteration 32600 (7.34348 iter/s, 27.2351s/200 iter), 67.8/104ep, loss = 0.0442505
  6138. I0818 14:56:45.657007 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0442505 (* 1 = 0.0442505 loss)
  6139. I0818 14:56:45.657045 18523 sgd_solver.cpp:180] [0.0] Iteration 32600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6140. I0818 14:56:54.227248 19072 data_reader.cpp:321] Restarting data pre-fetching
  6141. I0818 14:57:13.443614 18523 solver.cpp:333] [0.0] Iteration 32800 (7.19736 iter/s, 27.788s/200 iter), 68.2/104ep, loss = 0.0296173
  6142. I0818 14:57:13.443733 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0296173 (* 1 = 0.0296173 loss)
  6143. I0818 14:57:13.443771 18523 sgd_solver.cpp:180] [0.0] Iteration 32800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6144. I0818 14:57:40.345468 18523 solver.cpp:501] Iteration 33000, Testing net (#0)
  6145. I0818 14:57:46.198086 18511 data_reader.cpp:321] Restarting data pre-fetching
  6146. I0818 14:57:50.343546 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.90144
  6147. I0818 14:57:50.343612 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.318537 (* 1 = 0.318537 loss)
  6148. I0818 14:57:50.343701 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.8982s
  6149. I0818 14:57:50.475013 18523 solver.cpp:333] [0.0] Iteration 33000 (5.42032 iter/s, 36.8982s/200 iter), 68.6/104ep, loss = 0.161499
  6150. I0818 14:57:50.475129 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.161499 (* 1 = 0.161499 loss)
  6151. I0818 14:57:50.475145 18523 sgd_solver.cpp:180] [0.0] Iteration 33000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6152. I0818 14:58:09.633554 19072 data_reader.cpp:321] Restarting data pre-fetching
  6153. I0818 14:58:17.282980 18523 solver.cpp:333] [0.0] Iteration 33200 (7.46087 iter/s, 26.8065s/200 iter), 69.1/104ep, loss = 0.0184631
  6154. I0818 14:58:17.283324 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0184631 (* 1 = 0.0184631 loss)
  6155. I0818 14:58:17.283366 18523 sgd_solver.cpp:180] [0.0] Iteration 33200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6156. I0818 14:58:44.884981 18523 solver.cpp:333] [0.0] Iteration 33400 (7.24627 iter/s, 27.6004s/200 iter), 69.5/104ep, loss = 0.366699
  6157. I0818 14:58:44.885093 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.366699 (* 1 = 0.366699 loss)
  6158. I0818 14:58:44.885159 18523 sgd_solver.cpp:180] [0.0] Iteration 33400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6159. I0818 14:59:11.918792 18523 solver.cpp:333] [0.0] Iteration 33600 (7.39867 iter/s, 27.0319s/200 iter), 69.9/104ep, loss = 0.0291443
  6160. I0818 14:59:11.920141 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0291443 (* 1 = 0.0291443 loss)
  6161. I0818 14:59:11.920173 18523 sgd_solver.cpp:180] [0.0] Iteration 33600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6162. I0818 14:59:15.121170 19072 data_reader.cpp:321] Restarting data pre-fetching
  6163. I0818 14:59:39.038209 18523 solver.cpp:333] [0.0] Iteration 33800 (7.37528 iter/s, 27.1176s/200 iter), 70.3/104ep, loss = 0.0232697
  6164. I0818 14:59:39.038336 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0232697 (* 1 = 0.0232697 loss)
  6165. I0818 14:59:39.038381 18523 sgd_solver.cpp:180] [0.0] Iteration 33800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6166. I0818 15:00:05.658077 18523 solver.cpp:501] Iteration 34000, Testing net (#0)
  6167. I0818 15:00:11.161964 18511 data_reader.cpp:321] Restarting data pre-fetching
  6168. I0818 15:00:14.955905 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.89606
  6169. I0818 15:00:14.955942 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.330973 (* 1 = 0.330973 loss)
  6170. I0818 15:00:14.956034 18523 solver.cpp:271] [MultiGPU] Tests completed in 35.9154s
  6171. I0818 15:00:15.083840 18523 solver.cpp:333] [0.0] Iteration 34000 (5.56864 iter/s, 35.9154s/200 iter), 70.7/104ep, loss = 0.0150223
  6172. I0818 15:00:15.084023 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0150223 (* 1 = 0.0150223 loss)
  6173. I0818 15:00:15.084069 18523 sgd_solver.cpp:180] [0.0] Iteration 34000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6174. I0818 15:00:29.114495 19072 data_reader.cpp:321] Restarting data pre-fetching
  6175. I0818 15:00:41.549154 18523 solver.cpp:333] [0.0] Iteration 34200 (7.55757 iter/s, 26.4635s/200 iter), 71.1/104ep, loss = 0.00249672
  6176. I0818 15:00:41.552595 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00249672 (* 1 = 0.00249672 loss)
  6177. I0818 15:00:41.552613 18523 sgd_solver.cpp:180] [0.0] Iteration 34200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6178. I0818 15:01:08.318120 18523 solver.cpp:333] [0.0] Iteration 34400 (7.47186 iter/s, 26.7671s/200 iter), 71.6/104ep, loss = 0.00608063
  6179. I0818 15:01:08.318279 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00608063 (* 1 = 0.00608063 loss)
  6180. I0818 15:01:08.318316 18523 sgd_solver.cpp:180] [0.0] Iteration 34400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6181. I0818 15:01:33.505616 19072 data_reader.cpp:321] Restarting data pre-fetching
  6182. I0818 15:01:35.255712 18523 solver.cpp:333] [0.0] Iteration 34600 (7.42505 iter/s, 26.9359s/200 iter), 72/104ep, loss = 0.088562
  6183. I0818 15:01:35.255856 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.088562 (* 1 = 0.088562 loss)
  6184. I0818 15:01:35.255904 18523 sgd_solver.cpp:180] [0.0] Iteration 34600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6185. I0818 15:02:02.302994 18523 solver.cpp:333] [0.0] Iteration 34800 (7.39494 iter/s, 27.0455s/200 iter), 72.4/104ep, loss = 0.0567932
  6186. I0818 15:02:02.303169 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0567932 (* 1 = 0.0567932 loss)
  6187. I0818 15:02:02.303195 18523 sgd_solver.cpp:180] [0.0] Iteration 34800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6188. I0818 15:02:29.023937 18523 solver.cpp:501] Iteration 35000, Testing net (#0)
  6189. I0818 15:02:34.148033 18511 data_reader.cpp:321] Restarting data pre-fetching
  6190. I0818 15:02:38.643580 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.900664
  6191. I0818 15:02:38.643618 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.303727 (* 1 = 0.303727 loss)
  6192. I0818 15:02:38.643705 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.3384s
  6193. I0818 15:02:38.773245 18523 solver.cpp:333] [0.0] Iteration 35000 (5.50382 iter/s, 36.3384s/200 iter), 72.8/104ep, loss = 0.324219
  6194. I0818 15:02:38.773373 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.324219 (* 1 = 0.324219 loss)
  6195. I0818 15:02:38.773417 18523 sgd_solver.cpp:180] [0.0] Iteration 35000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6196. I0818 15:02:47.601217 19072 data_reader.cpp:321] Restarting data pre-fetching
  6197. I0818 15:03:05.326407 18523 solver.cpp:333] [0.0] Iteration 35200 (7.53256 iter/s, 26.5514s/200 iter), 73.2/104ep, loss = 0.0162659
  6198. I0818 15:03:05.326735 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0162659 (* 1 = 0.0162659 loss)
  6199. I0818 15:03:05.326758 18523 sgd_solver.cpp:180] [0.0] Iteration 35200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6200. I0818 15:03:32.361922 18523 solver.cpp:333] [0.0] Iteration 35400 (7.39814 iter/s, 27.0338s/200 iter), 73.6/104ep, loss = 0.0708618
  6201. I0818 15:03:32.362027 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0708618 (* 1 = 0.0708618 loss)
  6202. I0818 15:03:32.362056 18523 sgd_solver.cpp:180] [0.0] Iteration 35400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6203. I0818 15:03:52.244565 19072 data_reader.cpp:321] Restarting data pre-fetching
  6204. I0818 15:03:59.372439 18523 solver.cpp:333] [0.0] Iteration 35600 (7.40499 iter/s, 27.0088s/200 iter), 74/104ep, loss = 0.00261116
  6205. I0818 15:03:59.372534 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00261116 (* 1 = 0.00261116 loss)
  6206. I0818 15:03:59.372553 18523 sgd_solver.cpp:180] [0.0] Iteration 35600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6207. I0818 15:04:26.342818 18523 solver.cpp:333] [0.0] Iteration 35800 (7.41603 iter/s, 26.9686s/200 iter), 74.5/104ep, loss = 0.0830078
  6208. I0818 15:04:26.343168 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0830078 (* 1 = 0.0830078 loss)
  6209. I0818 15:04:26.343202 18523 sgd_solver.cpp:180] [0.0] Iteration 35800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6210. I0818 15:04:53.193522 18523 solver.cpp:501] Iteration 36000, Testing net (#0)
  6211. I0818 15:04:57.950834 18511 data_reader.cpp:321] Restarting data pre-fetching
  6212. I0818 15:05:02.625855 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.906055
  6213. I0818 15:05:02.625901 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.282692 (* 1 = 0.282692 loss)
  6214. I0818 15:05:02.625999 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.281s
  6215. I0818 15:05:02.755583 18523 solver.cpp:333] [0.0] Iteration 36000 (5.51253 iter/s, 36.281s/200 iter), 74.9/104ep, loss = 0.179321
  6216. I0818 15:05:02.755766 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.179321 (* 1 = 0.179321 loss)
  6217. I0818 15:05:02.755832 18523 sgd_solver.cpp:180] [0.0] Iteration 36000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6218. I0818 15:05:06.728538 19072 data_reader.cpp:321] Restarting data pre-fetching
  6219. I0818 15:05:29.387717 18523 solver.cpp:333] [0.0] Iteration 36200 (7.5102 iter/s, 26.6305s/200 iter), 75.3/104ep, loss = 0.0824585
  6220. I0818 15:05:29.388152 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0824585 (* 1 = 0.0824585 loss)
  6221. I0818 15:05:29.388198 18523 sgd_solver.cpp:180] [0.0] Iteration 36200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6222. I0818 15:05:56.645248 18523 solver.cpp:333] [0.0] Iteration 36400 (7.33787 iter/s, 27.2559s/200 iter), 75.7/104ep, loss = 0.0279083
  6223. I0818 15:05:56.645401 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0279083 (* 1 = 0.0279083 loss)
  6224. I0818 15:05:56.645438 18523 sgd_solver.cpp:180] [0.0] Iteration 36400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6225. I0818 15:06:11.425897 19072 data_reader.cpp:321] Restarting data pre-fetching
  6226. I0818 15:06:23.539659 18523 solver.cpp:333] [0.0] Iteration 36600 (7.43695 iter/s, 26.8927s/200 iter), 76.1/104ep, loss = 0.0413818
  6227. I0818 15:06:23.539772 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0413818 (* 1 = 0.0413818 loss)
  6228. I0818 15:06:23.539842 18523 sgd_solver.cpp:180] [0.0] Iteration 36600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6229. I0818 15:06:50.634809 18523 solver.cpp:333] [0.0] Iteration 36800 (7.38183 iter/s, 27.0935s/200 iter), 76.5/104ep, loss = 0.00124359
  6230. I0818 15:06:50.635177 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00124359 (* 1 = 0.00124359 loss)
  6231. I0818 15:06:50.635198 18523 sgd_solver.cpp:180] [0.0] Iteration 36800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6232. I0818 15:07:16.263670 19072 data_reader.cpp:321] Restarting data pre-fetching
  6233. I0818 15:07:17.458573 18523 solver.cpp:501] Iteration 37000, Testing net (#0)
  6234. I0818 15:07:21.854692 18511 data_reader.cpp:321] Restarting data pre-fetching
  6235. I0818 15:07:27.235816 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.913755
  6236. I0818 15:07:27.235873 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.276282 (* 1 = 0.276282 loss)
  6237. I0818 15:07:27.235972 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.599s
  6238. I0818 15:07:27.369199 18523 solver.cpp:333] [0.0] Iteration 37000 (5.46463 iter/s, 36.599s/200 iter), 77/104ep, loss = 0.00293922
  6239. I0818 15:07:27.369354 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00293922 (* 1 = 0.00293922 loss)
  6240. I0818 15:07:27.369385 18523 sgd_solver.cpp:180] [0.0] Iteration 37000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6241. I0818 15:07:54.045580 18523 solver.cpp:333] [0.0] Iteration 37200 (7.49774 iter/s, 26.6747s/200 iter), 77.4/104ep, loss = 0.0124969
  6242. I0818 15:07:54.045912 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0124969 (* 1 = 0.0124969 loss)
  6243. I0818 15:07:54.045951 18523 sgd_solver.cpp:180] [0.0] Iteration 37200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6244. I0818 15:08:21.355824 18523 solver.cpp:333] [0.0] Iteration 37400 (7.32369 iter/s, 27.3086s/200 iter), 77.8/104ep, loss = 0.00665283
  6245. I0818 15:08:21.355916 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00665283 (* 1 = 0.00665283 loss)
  6246. I0818 15:08:21.355945 18523 sgd_solver.cpp:180] [0.0] Iteration 37400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6247. I0818 15:08:30.894541 19072 data_reader.cpp:321] Restarting data pre-fetching
  6248. I0818 15:08:48.141392 18523 solver.cpp:333] [0.0] Iteration 37600 (7.46716 iter/s, 26.784s/200 iter), 78.2/104ep, loss = 0.256348
  6249. I0818 15:08:48.141674 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.256348 (* 1 = 0.256348 loss)
  6250. I0818 15:08:48.141713 18523 sgd_solver.cpp:180] [0.0] Iteration 37600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6251. I0818 15:09:15.311863 18523 solver.cpp:333] [0.0] Iteration 37800 (7.36137 iter/s, 27.1689s/200 iter), 78.6/104ep, loss = 0.0168915
  6252. I0818 15:09:15.312130 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0168915 (* 1 = 0.0168915 loss)
  6253. I0818 15:09:15.312173 18523 sgd_solver.cpp:180] [0.0] Iteration 37800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6254. I0818 15:09:35.807546 19072 data_reader.cpp:321] Restarting data pre-fetching
  6255. I0818 15:09:42.055599 18523 solver.cpp:501] Iteration 38000, Testing net (#0)
  6256. I0818 15:09:46.103147 18511 data_reader.cpp:321] Restarting data pre-fetching
  6257. I0818 15:09:46.188694 18823 blocking_queue.cpp:40] Data layer prefetch queue empty
  6258. I0818 15:09:51.702214 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.912993
  6259. I0818 15:09:51.702266 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.293364 (* 1 = 0.293364 loss)
  6260. I0818 15:09:51.702324 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.3884s
  6261. I0818 15:09:51.837211 18523 solver.cpp:333] [0.0] Iteration 38000 (5.49626 iter/s, 36.3884s/200 iter), 79/104ep, loss = 0.00219536
  6262. I0818 15:09:51.837381 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00219536 (* 1 = 0.00219536 loss)
  6263. I0818 15:09:51.837419 18523 sgd_solver.cpp:180] [0.0] Iteration 38000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6264. I0818 15:10:19.023047 18523 solver.cpp:333] [0.0] Iteration 38200 (7.35721 iter/s, 27.1842s/200 iter), 79.5/104ep, loss = 0.109436
  6265. I0818 15:10:19.023401 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.109436 (* 1 = 0.109436 loss)
  6266. I0818 15:10:19.023447 18523 sgd_solver.cpp:180] [0.0] Iteration 38200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6267. I0818 15:10:46.381618 18523 solver.cpp:333] [0.0] Iteration 38400 (7.31076 iter/s, 27.3569s/200 iter), 79.9/104ep, loss = 0.0174713
  6268. I0818 15:10:46.381791 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0174713 (* 1 = 0.0174713 loss)
  6269. I0818 15:10:46.381834 18523 sgd_solver.cpp:180] [0.0] Iteration 38400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6270. I0818 15:10:50.998895 19072 data_reader.cpp:321] Restarting data pre-fetching
  6271. I0818 15:11:13.493532 18523 solver.cpp:333] [0.0] Iteration 38600 (7.37725 iter/s, 27.1104s/200 iter), 80.3/104ep, loss = 0.0265961
  6272. I0818 15:11:13.493681 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0265961 (* 1 = 0.0265961 loss)
  6273. I0818 15:11:13.493741 18523 sgd_solver.cpp:180] [0.0] Iteration 38600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6274. I0818 15:11:40.932082 18523 solver.cpp:333] [0.0] Iteration 38800 (7.28945 iter/s, 27.4369s/200 iter), 80.7/104ep, loss = 0.0877075
  6275. I0818 15:11:40.932296 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0877075 (* 1 = 0.0877075 loss)
  6276. I0818 15:11:40.932335 18523 sgd_solver.cpp:180] [0.0] Iteration 38800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6277. I0818 15:11:56.247267 19072 data_reader.cpp:321] Restarting data pre-fetching
  6278. I0818 15:12:07.970250 18523 solver.cpp:501] Iteration 39000, Testing net (#0)
  6279. I0818 15:12:11.743973 18511 data_reader.cpp:321] Restarting data pre-fetching
  6280. I0818 15:12:17.348824 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.916069
  6281. I0818 15:12:17.348881 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.274479 (* 1 = 0.274479 loss)
  6282. I0818 15:12:17.348963 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.4148s
  6283. I0818 15:12:17.483441 18523 solver.cpp:333] [0.0] Iteration 39000 (5.49227 iter/s, 36.4148s/200 iter), 81.1/104ep, loss = 0.0102997
  6284. I0818 15:12:17.483630 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0102997 (* 1 = 0.0102997 loss)
  6285. I0818 15:12:17.483657 18523 sgd_solver.cpp:180] [0.0] Iteration 39000, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6286. I0818 15:12:44.478354 18523 solver.cpp:333] [0.0] Iteration 39200 (7.40923 iter/s, 26.9933s/200 iter), 81.5/104ep, loss = 0.0198669
  6287. I0818 15:12:44.480096 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0198669 (* 1 = 0.0198669 loss)
  6288. I0818 15:12:44.480115 18523 sgd_solver.cpp:180] [0.0] Iteration 39200, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6289. I0818 15:13:10.975227 19072 data_reader.cpp:321] Restarting data pre-fetching
  6290. I0818 15:13:11.499047 18523 solver.cpp:333] [0.0] Iteration 39400 (7.40217 iter/s, 27.0191s/200 iter), 82/104ep, loss = 0.0313416
  6291. I0818 15:13:11.499132 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0313416 (* 1 = 0.0313416 loss)
  6292. I0818 15:13:11.499146 18523 sgd_solver.cpp:180] [0.0] Iteration 39400, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6293. I0818 15:13:39.070734 18523 solver.cpp:333] [0.0] Iteration 39600 (7.25422 iter/s, 27.5701s/200 iter), 82.4/104ep, loss = 0.00400543
  6294. I0818 15:13:39.071238 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00400543 (* 1 = 0.00400543 loss)
  6295. I0818 15:13:39.071262 18523 sgd_solver.cpp:180] [0.0] Iteration 39600, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6296. I0818 15:14:06.431077 18523 solver.cpp:333] [0.0] Iteration 39800 (7.31027 iter/s, 27.3588s/200 iter), 82.8/104ep, loss = 0.00562286
  6297. I0818 15:14:06.431170 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00562286 (* 1 = 0.00562286 loss)
  6298. I0818 15:14:06.431197 18523 sgd_solver.cpp:180] [0.0] Iteration 39800, lr = 0.0001, m = 0.9, lrm = 0.001, wd = 0.004, gs = 2000
  6299. I0818 15:14:16.805675 19072 data_reader.cpp:321] Restarting data pre-fetching
  6300. I0818 15:14:33.616510 18523 solver.cpp:501] Iteration 40000, Testing net (#0)
  6301. I0818 15:14:37.031877 18511 data_reader.cpp:321] Restarting data pre-fetching
  6302. I0818 15:14:42.988441 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.895298
  6303. I0818 15:14:42.988499 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.354747 (* 1 = 0.354747 loss)
  6304. I0818 15:14:42.988572 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.5555s
  6305. I0818 15:14:42.994451 19005 sgd_solver.cpp:50] MultiStep Status: Iteration 40000, step = 2
  6306. I0818 15:14:42.994760 19009 sgd_solver.cpp:50] MultiStep Status: Iteration 40000, step = 2
  6307. I0818 15:14:42.994940 19004 sgd_solver.cpp:50] MultiStep Status: Iteration 40000, step = 2
  6308. I0818 15:14:42.995149 19010 sgd_solver.cpp:50] MultiStep Status: Iteration 40000, step = 2
  6309. I0818 15:14:42.995209 19006 sgd_solver.cpp:50] MultiStep Status: Iteration 40000, step = 2
  6310. I0818 15:14:42.995302 19008 sgd_solver.cpp:50] MultiStep Status: Iteration 40000, step = 2
  6311. I0818 15:14:42.995323 19011 sgd_solver.cpp:50] MultiStep Status: Iteration 40000, step = 2
  6312. I0818 15:14:42.995378 19007 sgd_solver.cpp:50] MultiStep Status: Iteration 40000, step = 2
  6313. I0818 15:14:43.119704 18523 solver.cpp:333] [0.0] Iteration 40000 (5.47114 iter/s, 36.5555s/200 iter), 83.2/104ep, loss = 0.00741959
  6314. I0818 15:14:43.119948 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00741959 (* 1 = 0.00741959 loss)
  6315. I0818 15:14:43.119988 18523 sgd_solver.cpp:180] [0.0] Iteration 40000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6316. I0818 15:15:10.416901 18523 solver.cpp:333] [0.0] Iteration 40200 (7.32719 iter/s, 27.2956s/200 iter), 83.6/104ep, loss = 0.00105572
  6317. I0818 15:15:10.417203 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00105572 (* 1 = 0.00105572 loss)
  6318. I0818 15:15:10.417273 18523 sgd_solver.cpp:180] [0.0] Iteration 40200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6319. I0818 15:15:31.859706 19072 data_reader.cpp:321] Restarting data pre-fetching
  6320. I0818 15:15:37.953919 18523 solver.cpp:333] [0.0] Iteration 40400 (7.26336 iter/s, 27.5355s/200 iter), 84/104ep, loss = 0.0155411
  6321. I0818 15:15:37.954169 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0155411 (* 1 = 0.0155411 loss)
  6322. I0818 15:15:37.954231 18523 sgd_solver.cpp:180] [0.0] Iteration 40400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6323. I0818 15:16:05.701725 18523 solver.cpp:333] [0.0] Iteration 40600 (7.20819 iter/s, 27.7462s/200 iter), 84.4/104ep, loss = 0.0293884
  6324. I0818 15:16:05.704399 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0293884 (* 1 = 0.0293884 loss)
  6325. I0818 15:16:05.704432 18523 sgd_solver.cpp:180] [0.0] Iteration 40600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6326. I0818 15:16:33.326483 18523 solver.cpp:333] [0.0] Iteration 40800 (7.24028 iter/s, 27.6232s/200 iter), 84.9/104ep, loss = 0.00501633
  6327. I0818 15:16:33.326611 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00501633 (* 1 = 0.00501633 loss)
  6328. I0818 15:16:33.326709 18523 sgd_solver.cpp:180] [0.0] Iteration 40800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6329. I0818 15:16:38.297999 19072 data_reader.cpp:321] Restarting data pre-fetching
  6330. I0818 15:17:00.684026 18523 solver.cpp:501] Iteration 41000, Testing net (#0)
  6331. I0818 15:17:03.740926 18511 data_reader.cpp:321] Restarting data pre-fetching
  6332. I0818 15:17:10.141866 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.91145
  6333. I0818 15:17:10.142040 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.291586 (* 1 = 0.291586 loss)
  6334. I0818 15:17:10.142127 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.8136s
  6335. I0818 15:17:10.274508 18523 solver.cpp:333] [0.0] Iteration 41000 (5.43277 iter/s, 36.8136s/200 iter), 85.3/104ep, loss = 0.00241661
  6336. I0818 15:17:10.274595 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00241661 (* 1 = 0.00241661 loss)
  6337. I0818 15:17:10.274639 18523 sgd_solver.cpp:180] [0.0] Iteration 41000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6338. I0818 15:17:37.531508 18523 solver.cpp:333] [0.0] Iteration 41200 (7.33798 iter/s, 27.2555s/200 iter), 85.7/104ep, loss = 0.0147324
  6339. I0818 15:17:37.531607 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0147324 (* 1 = 0.0147324 loss)
  6340. I0818 15:17:37.531625 18523 sgd_solver.cpp:180] [0.0] Iteration 41200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6341. I0818 15:17:53.870040 19072 data_reader.cpp:321] Restarting data pre-fetching
  6342. I0818 15:18:04.942478 18523 solver.cpp:333] [0.0] Iteration 41400 (7.29678 iter/s, 27.4094s/200 iter), 86.1/104ep, loss = 0.217163
  6343. I0818 15:18:04.942569 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.217163 (* 1 = 0.217163 loss)
  6344. I0818 15:18:04.942584 18523 sgd_solver.cpp:180] [0.0] Iteration 41400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6345. I0818 15:18:32.809640 18523 solver.cpp:333] [0.0] Iteration 41600 (7.17731 iter/s, 27.8656s/200 iter), 86.5/104ep, loss = 0.0765381
  6346. I0818 15:18:32.809991 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0765381 (* 1 = 0.0765381 loss)
  6347. I0818 15:18:32.810045 18523 sgd_solver.cpp:180] [0.0] Iteration 41600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6348. I0818 15:19:00.498709 19072 data_reader.cpp:321] Restarting data pre-fetching
  6349. I0818 15:19:00.633474 18523 solver.cpp:333] [0.0] Iteration 41800 (7.18849 iter/s, 27.8223s/200 iter), 86.9/104ep, loss = 0.00410461
  6350. I0818 15:19:00.633616 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00410461 (* 1 = 0.00410461 loss)
  6351. I0818 15:19:00.633646 18523 sgd_solver.cpp:180] [0.0] Iteration 41800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6352. I0818 15:19:28.076269 18523 solver.cpp:501] Iteration 42000, Testing net (#0)
  6353. I0818 15:19:30.773402 18511 data_reader.cpp:321] Restarting data pre-fetching
  6354. I0818 15:19:37.562734 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.910649
  6355. I0818 15:19:37.562788 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.258106 (* 1 = 0.258106 loss)
  6356. I0818 15:19:37.562887 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.9274s
  6357. I0818 15:19:37.696230 18523 solver.cpp:333] [0.0] Iteration 42000 (5.41603 iter/s, 36.9274s/200 iter), 87.4/104ep, loss = 0.0587463
  6358. I0818 15:19:37.696441 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0587463 (* 1 = 0.0587463 loss)
  6359. I0818 15:19:37.696470 18523 sgd_solver.cpp:180] [0.0] Iteration 42000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6360. I0818 15:20:05.060459 18523 solver.cpp:333] [0.0] Iteration 42200 (7.30922 iter/s, 27.3627s/200 iter), 87.8/104ep, loss = 0.00177765
  6361. I0818 15:20:05.062240 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00177765 (* 1 = 0.00177765 loss)
  6362. I0818 15:20:05.062259 18523 sgd_solver.cpp:180] [0.0] Iteration 42200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6363. I0818 15:20:15.902423 19072 data_reader.cpp:321] Restarting data pre-fetching
  6364. I0818 15:20:32.589210 18523 solver.cpp:333] [0.0] Iteration 42400 (7.26553 iter/s, 27.5272s/200 iter), 88.2/104ep, loss = 0.000338554
  6365. I0818 15:20:32.589357 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.000338554 (* 1 = 0.000338554 loss)
  6366. I0818 15:20:32.589416 18523 sgd_solver.cpp:180] [0.0] Iteration 42400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6367. I0818 15:21:00.428812 18523 solver.cpp:333] [0.0] Iteration 42600 (7.18442 iter/s, 27.838s/200 iter), 88.6/104ep, loss = 0.00313568
  6368. I0818 15:21:00.429168 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00313568 (* 1 = 0.00313568 loss)
  6369. I0818 15:21:00.429217 18523 sgd_solver.cpp:180] [0.0] Iteration 42600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6370. I0818 15:21:22.474474 19072 data_reader.cpp:321] Restarting data pre-fetching
  6371. I0818 15:21:28.203161 18523 solver.cpp:333] [0.0] Iteration 42800 (7.20128 iter/s, 27.7728s/200 iter), 89/104ep, loss = 0.0338135
  6372. I0818 15:21:28.203306 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0338135 (* 1 = 0.0338135 loss)
  6373. I0818 15:21:28.203354 18523 sgd_solver.cpp:180] [0.0] Iteration 42800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6374. I0818 15:21:55.263805 18523 solver.cpp:501] Iteration 43000, Testing net (#0)
  6375. I0818 15:21:57.613021 18511 data_reader.cpp:321] Restarting data pre-fetching
  6376. I0818 15:22:04.937783 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.90915
  6377. I0818 15:22:04.937841 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.298344 (* 1 = 0.298344 loss)
  6378. I0818 15:22:04.937949 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.7328s
  6379. I0818 15:22:05.069137 18523 solver.cpp:333] [0.0] Iteration 43000 (5.44472 iter/s, 36.7328s/200 iter), 89.4/104ep, loss = 0.345703
  6380. I0818 15:22:05.069413 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.345703 (* 1 = 0.345703 loss)
  6381. I0818 15:22:05.069522 18523 sgd_solver.cpp:180] [0.0] Iteration 43000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6382. I0818 15:22:31.970144 18523 solver.cpp:333] [0.0] Iteration 43200 (7.43508 iter/s, 26.8995s/200 iter), 89.9/104ep, loss = 0.0838623
  6383. I0818 15:22:31.970410 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0838623 (* 1 = 0.0838623 loss)
  6384. I0818 15:22:31.970427 18523 sgd_solver.cpp:180] [0.0] Iteration 43200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6385. I0818 15:22:37.654742 19072 data_reader.cpp:321] Restarting data pre-fetching
  6386. I0818 15:22:59.767802 18523 solver.cpp:333] [0.0] Iteration 43400 (7.19526 iter/s, 27.7961s/200 iter), 90.3/104ep, loss = 0.0515747
  6387. I0818 15:22:59.767962 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0515747 (* 1 = 0.0515747 loss)
  6388. I0818 15:22:59.768009 18523 sgd_solver.cpp:180] [0.0] Iteration 43400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6389. I0818 15:23:27.005398 18523 solver.cpp:333] [0.0] Iteration 43600 (7.3432 iter/s, 27.2361s/200 iter), 90.7/104ep, loss = 0.0187073
  6390. I0818 15:23:27.005924 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0187073 (* 1 = 0.0187073 loss)
  6391. I0818 15:23:27.005991 18523 sgd_solver.cpp:180] [0.0] Iteration 43600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6392. I0818 15:23:43.606663 19072 data_reader.cpp:321] Restarting data pre-fetching
  6393. I0818 15:23:54.409719 18523 solver.cpp:333] [0.0] Iteration 43800 (7.29852 iter/s, 27.4028s/200 iter), 91.1/104ep, loss = 0.00439072
  6394. I0818 15:23:54.409878 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00439072 (* 1 = 0.00439072 loss)
  6395. I0818 15:23:54.409919 18523 sgd_solver.cpp:180] [0.0] Iteration 43800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6396. I0818 15:24:21.776949 18523 solver.cpp:501] Iteration 44000, Testing net (#0)
  6397. I0818 15:24:23.761924 18511 data_reader.cpp:321] Restarting data pre-fetching
  6398. I0818 15:24:31.147575 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.918364
  6399. I0818 15:24:31.147634 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.279126 (* 1 = 0.279126 loss)
  6400. I0818 15:24:31.147735 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.736s
  6401. I0818 15:24:31.278769 18523 solver.cpp:333] [0.0] Iteration 44000 (5.44425 iter/s, 36.736s/200 iter), 91.5/104ep, loss = 0.0267029
  6402. I0818 15:24:31.278954 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0267029 (* 1 = 0.0267029 loss)
  6403. I0818 15:24:31.279000 18523 sgd_solver.cpp:180] [0.0] Iteration 44000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6404. I0818 15:24:58.286979 18523 solver.cpp:333] [0.0] Iteration 44200 (7.40559 iter/s, 27.0066s/200 iter), 91.9/104ep, loss = 0.0009408
  6405. I0818 15:24:58.287467 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0009408 (* 1 = 0.0009408 loss)
  6406. I0818 15:24:58.287487 18523 sgd_solver.cpp:180] [0.0] Iteration 44200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6407. I0818 15:24:58.561566 19072 data_reader.cpp:321] Restarting data pre-fetching
  6408. I0818 15:25:25.875718 18523 solver.cpp:333] [0.0] Iteration 44400 (7.24972 iter/s, 27.5873s/200 iter), 92.4/104ep, loss = 0.0266418
  6409. I0818 15:25:25.875916 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0266418 (* 1 = 0.0266418 loss)
  6410. I0818 15:25:25.875989 18523 sgd_solver.cpp:180] [0.0] Iteration 44400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6411. I0818 15:25:53.369521 18523 solver.cpp:333] [0.0] Iteration 44600 (7.27477 iter/s, 27.4923s/200 iter), 92.8/104ep, loss = 0.015007
  6412. I0818 15:25:53.369769 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.015007 (* 1 = 0.015007 loss)
  6413. I0818 15:25:53.369805 18523 sgd_solver.cpp:180] [0.0] Iteration 44600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6414. I0818 15:26:04.567487 19072 data_reader.cpp:321] Restarting data pre-fetching
  6415. I0818 15:26:20.798955 18523 solver.cpp:333] [0.0] Iteration 44800 (7.29185 iter/s, 27.4279s/200 iter), 93.2/104ep, loss = 0.167358
  6416. I0818 15:26:20.799120 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.167358 (* 1 = 0.167358 loss)
  6417. I0818 15:26:20.799176 18523 sgd_solver.cpp:180] [0.0] Iteration 44800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6418. I0818 15:26:48.279479 18523 solver.cpp:501] Iteration 45000, Testing net (#0)
  6419. I0818 15:26:49.992089 18511 data_reader.cpp:321] Restarting data pre-fetching
  6420. I0818 15:26:57.808585 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.90835
  6421. I0818 15:26:57.808624 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.292754 (* 1 = 0.292754 loss)
  6422. I0818 15:26:57.808693 18523 solver.cpp:271] [MultiGPU] Tests completed in 37.0078s
  6423. I0818 15:26:57.943624 18523 solver.cpp:333] [0.0] Iteration 45000 (5.40426 iter/s, 37.0078s/200 iter), 93.6/104ep, loss = 0.0290985
  6424. I0818 15:26:57.943737 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0290985 (* 1 = 0.0290985 loss)
  6425. I0818 15:26:57.943790 18523 sgd_solver.cpp:180] [0.0] Iteration 45000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6426. I0818 15:27:20.503343 19072 data_reader.cpp:321] Restarting data pre-fetching
  6427. I0818 15:27:25.365370 18523 solver.cpp:333] [0.0] Iteration 45200 (7.2939 iter/s, 27.4202s/200 iter), 94/104ep, loss = 0.0232697
  6428. I0818 15:27:25.365624 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0232697 (* 1 = 0.0232697 loss)
  6429. I0818 15:27:25.365700 18523 sgd_solver.cpp:180] [0.0] Iteration 45200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6430. I0818 15:27:52.816555 18523 solver.cpp:333] [0.0] Iteration 45400 (7.28605 iter/s, 27.4497s/200 iter), 94.4/104ep, loss = 0.159058
  6431. I0818 15:27:52.816833 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.159058 (* 1 = 0.159058 loss)
  6432. I0818 15:27:52.816874 18523 sgd_solver.cpp:180] [0.0] Iteration 45400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6433. I0818 15:28:20.249155 18523 solver.cpp:333] [0.0] Iteration 45600 (7.291 iter/s, 27.4311s/200 iter), 94.8/104ep, loss = 0.0387573
  6434. I0818 15:28:20.249296 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0387573 (* 1 = 0.0387573 loss)
  6435. I0818 15:28:20.249341 18523 sgd_solver.cpp:180] [0.0] Iteration 45600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6436. I0818 15:28:26.394573 19072 data_reader.cpp:321] Restarting data pre-fetching
  6437. I0818 15:28:47.862356 18523 solver.cpp:333] [0.0] Iteration 45800 (7.2433 iter/s, 27.6117s/200 iter), 95.3/104ep, loss = 0.150513
  6438. I0818 15:28:47.862525 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.150513 (* 1 = 0.150513 loss)
  6439. I0818 15:28:47.862568 18523 sgd_solver.cpp:180] [0.0] Iteration 45800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6440. I0818 15:29:15.434232 18523 solver.cpp:501] Iteration 46000, Testing net (#0)
  6441. I0818 15:29:16.795461 18511 data_reader.cpp:321] Restarting data pre-fetching
  6442. I0818 15:29:24.862484 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.90145
  6443. I0818 15:29:24.862522 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.310386 (* 1 = 0.310386 loss)
  6444. I0818 15:29:24.862604 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.9983s
  6445. I0818 15:29:24.993930 18523 solver.cpp:333] [0.0] Iteration 46000 (5.40565 iter/s, 36.9983s/200 iter), 95.7/104ep, loss = 0.189575
  6446. I0818 15:29:24.994153 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.189575 (* 1 = 0.189575 loss)
  6447. I0818 15:29:24.994207 18523 sgd_solver.cpp:180] [0.0] Iteration 46000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6448. I0818 15:29:41.872107 19072 data_reader.cpp:321] Restarting data pre-fetching
  6449. I0818 15:29:52.171660 18523 solver.cpp:333] [0.0] Iteration 46200 (7.35938 iter/s, 27.1762s/200 iter), 96.1/104ep, loss = 0.0651245
  6450. I0818 15:29:52.171922 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0651245 (* 1 = 0.0651245 loss)
  6451. I0818 15:29:52.171981 18523 sgd_solver.cpp:180] [0.0] Iteration 46200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6452. I0818 15:30:19.924808 18523 solver.cpp:333] [0.0] Iteration 46400 (7.20678 iter/s, 27.7516s/200 iter), 96.5/104ep, loss = 0.267822
  6453. I0818 15:30:19.925015 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.267822 (* 1 = 0.267822 loss)
  6454. I0818 15:30:19.925057 18523 sgd_solver.cpp:180] [0.0] Iteration 46400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6455. I0818 15:30:47.514168 18523 solver.cpp:333] [0.0] Iteration 46600 (7.24958 iter/s, 27.5878s/200 iter), 96.9/104ep, loss = 0.0741577
  6456. I0818 15:30:47.514451 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0741577 (* 1 = 0.0741577 loss)
  6457. I0818 15:30:47.514519 18523 sgd_solver.cpp:180] [0.0] Iteration 46600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6458. I0818 15:30:48.209977 19072 data_reader.cpp:321] Restarting data pre-fetching
  6459. I0818 15:31:15.256618 18523 solver.cpp:333] [0.0] Iteration 46800 (7.20958 iter/s, 27.7409s/200 iter), 97.3/104ep, loss = 0.193237
  6460. I0818 15:31:15.256814 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.193237 (* 1 = 0.193237 loss)
  6461. I0818 15:31:15.256851 18523 sgd_solver.cpp:180] [0.0] Iteration 46800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6462. I0818 15:31:42.391002 18523 solver.cpp:501] Iteration 47000, Testing net (#0)
  6463. I0818 15:31:43.400806 18511 data_reader.cpp:321] Restarting data pre-fetching
  6464. I0818 15:31:52.078948 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.904526
  6465. I0818 15:31:52.079003 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.308453 (* 1 = 0.308453 loss)
  6466. I0818 15:31:52.079080 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.8206s
  6467. I0818 15:31:52.211243 18523 solver.cpp:333] [0.0] Iteration 47000 (5.43174 iter/s, 36.8206s/200 iter), 97.8/104ep, loss = 0.270752
  6468. I0818 15:31:52.211421 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.270752 (* 1 = 0.270752 loss)
  6469. I0818 15:31:52.211482 18523 sgd_solver.cpp:180] [0.0] Iteration 47000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6470. I0818 15:32:04.005053 19072 data_reader.cpp:321] Restarting data pre-fetching
  6471. I0818 15:32:19.150255 18523 solver.cpp:333] [0.0] Iteration 47200 (7.4246 iter/s, 26.9375s/200 iter), 98.2/104ep, loss = 0.0686646
  6472. I0818 15:32:19.150599 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0686646 (* 1 = 0.0686646 loss)
  6473. I0818 15:32:19.150632 18523 sgd_solver.cpp:180] [0.0] Iteration 47200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6474. I0818 15:32:47.134312 18523 solver.cpp:333] [0.0] Iteration 47400 (7.1473 iter/s, 27.9826s/200 iter), 98.6/104ep, loss = 0.109985
  6475. I0818 15:32:47.134452 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.109985 (* 1 = 0.109985 loss)
  6476. I0818 15:32:47.134491 18523 sgd_solver.cpp:180] [0.0] Iteration 47400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6477. I0818 15:33:10.284879 19072 data_reader.cpp:321] Restarting data pre-fetching
  6478. I0818 15:33:14.781335 18523 solver.cpp:333] [0.0] Iteration 47600 (7.23441 iter/s, 27.6457s/200 iter), 99/104ep, loss = 0.0389709
  6479. I0818 15:33:14.781463 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0389709 (* 1 = 0.0389709 loss)
  6480. I0818 15:33:14.781488 18523 sgd_solver.cpp:180] [0.0] Iteration 47600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6481. I0818 15:33:42.184155 18523 solver.cpp:333] [0.0] Iteration 47800 (7.29888 iter/s, 27.4015s/200 iter), 99.4/104ep, loss = 0.281738
  6482. I0818 15:33:42.184343 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.281738 (* 1 = 0.281738 loss)
  6483. I0818 15:33:42.184365 18523 sgd_solver.cpp:180] [0.0] Iteration 47800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6484. I0818 15:34:09.315850 18523 solver.cpp:501] Iteration 48000, Testing net (#0)
  6485. I0818 15:34:09.951653 18511 data_reader.cpp:321] Restarting data pre-fetching
  6486. I0818 15:34:18.922449 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.922983
  6487. I0818 15:34:18.923856 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.242856 (* 1 = 0.242856 loss)
  6488. I0818 15:34:18.923959 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.738s
  6489. I0818 15:34:19.056156 18523 solver.cpp:333] [0.0] Iteration 48000 (5.44395 iter/s, 36.738s/200 iter), 99.8/104ep, loss = 0.166016
  6490. I0818 15:34:19.056404 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.166016 (* 1 = 0.166016 loss)
  6491. I0818 15:34:19.056448 18523 sgd_solver.cpp:180] [0.0] Iteration 48000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6492. I0818 15:34:25.522579 19072 data_reader.cpp:321] Restarting data pre-fetching
  6493. I0818 15:34:46.424271 18523 solver.cpp:333] [0.0] Iteration 48200 (7.30814 iter/s, 27.3667s/200 iter), 100.3/104ep, loss = 0.0570068
  6494. I0818 15:34:46.424414 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0570068 (* 1 = 0.0570068 loss)
  6495. I0818 15:34:46.424438 18523 sgd_solver.cpp:180] [0.0] Iteration 48200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6496. I0818 15:35:13.799773 18523 solver.cpp:333] [0.0] Iteration 48400 (7.30618 iter/s, 27.3741s/200 iter), 100.7/104ep, loss = 0.0541077
  6497. I0818 15:35:13.800025 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0541077 (* 1 = 0.0541077 loss)
  6498. I0818 15:35:13.800045 18523 sgd_solver.cpp:180] [0.0] Iteration 48400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6499. I0818 15:35:31.343917 19072 data_reader.cpp:321] Restarting data pre-fetching
  6500. I0818 15:35:41.193006 18523 solver.cpp:333] [0.0] Iteration 48600 (7.30143 iter/s, 27.3919s/200 iter), 101.1/104ep, loss = 0.122925
  6501. I0818 15:35:41.193111 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.122925 (* 1 = 0.122925 loss)
  6502. I0818 15:35:41.193136 18523 sgd_solver.cpp:180] [0.0] Iteration 48600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6503. I0818 15:36:08.726785 18523 solver.cpp:333] [0.0] Iteration 48800 (7.26417 iter/s, 27.5324s/200 iter), 101.5/104ep, loss = 0.0226898
  6504. I0818 15:36:08.727061 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0226898 (* 1 = 0.0226898 loss)
  6505. I0818 15:36:08.727079 18523 sgd_solver.cpp:180] [0.0] Iteration 48800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6506. I0818 15:36:35.573521 18523 solver.cpp:501] Iteration 49000, Testing net (#0)
  6507. I0818 15:36:35.854141 18511 data_reader.cpp:321] Restarting data pre-fetching
  6508. I0818 15:36:44.459930 18511 data_reader.cpp:321] Restarting data pre-fetching
  6509. I0818 15:36:45.563992 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.903745
  6510. I0818 15:36:45.564047 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.298249 (* 1 = 0.298249 loss)
  6511. I0818 15:36:45.564117 18523 solver.cpp:271] [MultiGPU] Tests completed in 36.8355s
  6512. I0818 15:36:45.700634 18523 solver.cpp:333] [0.0] Iteration 49000 (5.42955 iter/s, 36.8355s/200 iter), 101.9/104ep, loss = 0.0542603
  6513. I0818 15:36:45.700894 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0542603 (* 1 = 0.0542603 loss)
  6514. I0818 15:36:45.700927 18523 sgd_solver.cpp:180] [0.0] Iteration 49000, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6515. I0818 15:36:47.188035 19072 data_reader.cpp:321] Restarting data pre-fetching
  6516. I0818 15:37:13.199812 18523 solver.cpp:333] [0.0] Iteration 49200 (7.27331 iter/s, 27.4978s/200 iter), 102.3/104ep, loss = 0.0170288
  6517. I0818 15:37:13.199905 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0170288 (* 1 = 0.0170288 loss)
  6518. I0818 15:37:13.199932 18523 sgd_solver.cpp:180] [0.0] Iteration 49200, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6519. I0818 15:37:40.602689 18523 solver.cpp:333] [0.0] Iteration 49400 (7.29889 iter/s, 27.4014s/200 iter), 102.8/104ep, loss = 0.0167847
  6520. I0818 15:37:40.603096 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0167847 (* 1 = 0.0167847 loss)
  6521. I0818 15:37:40.603123 18523 sgd_solver.cpp:180] [0.0] Iteration 49400, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6522. I0818 15:37:53.129655 19072 data_reader.cpp:321] Restarting data pre-fetching
  6523. I0818 15:38:08.082723 18523 solver.cpp:333] [0.0] Iteration 49600 (7.27838 iter/s, 27.4786s/200 iter), 103.2/104ep, loss = 0.00336838
  6524. I0818 15:38:08.082849 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.00336838 (* 1 = 0.00336838 loss)
  6525. I0818 15:38:08.082901 18523 sgd_solver.cpp:180] [0.0] Iteration 49600, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6526. I0818 15:38:35.617394 18523 solver.cpp:333] [0.0] Iteration 49800 (7.26393 iter/s, 27.5333s/200 iter), 103.6/104ep, loss = 0.0615845
  6527. I0818 15:38:35.617709 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0615845 (* 1 = 0.0615845 loss)
  6528. I0818 15:38:35.617848 18523 sgd_solver.cpp:180] [0.0] Iteration 49800, lr = 1e-05, m = 0.9, lrm = 0.0001, wd = 0.004, gs = 2000
  6529. I0818 15:38:59.044800 19072 data_reader.cpp:321] Restarting data pre-fetching
  6530. I0818 15:39:02.864567 18523 solver.cpp:333] [0.0] Iteration 50000 (7.30389 iter/s, 27.2457s/199 iter), 104/104ep, loss = 0.0231781
  6531. I0818 15:39:02.864665 18523 solver.cpp:361] [0.0] Train net output #0: loss = 0.0231781 (* 1 = 0.0231781 loss)
  6532. I0818 15:39:02.867518 18523 solver.cpp:769] Snapshotting to binary proto file _iter_50000.caffemodel
  6533. I0818 15:39:03.011508 18523 sgd_solver.cpp:448] Snapshotting solver state to binary proto file _iter_50000.solverstate
  6534. I0818 15:39:03.076583 18523 solver.cpp:466] Iteration 50000, loss = 0.0248718
  6535. I0818 15:39:03.076632 18523 solver.cpp:501] Iteration 50000, Testing net (#0)
  6536. I0818 15:39:11.591514 18510 blocking_queue.cpp:40] Data layer prefetch queue empty
  6537. I0818 15:39:11.591715 18511 data_reader.cpp:321] Restarting data pre-fetching
  6538. I0818 15:39:12.518956 18523 solver.cpp:588] (0.0) Test net output #0: accuracy = 0.920684
  6539. I0818 15:39:12.519008 18523 solver.cpp:588] (0.0) Test net output #1: loss = 0.264779 (* 1 = 0.264779 loss)
  6540. I0818 15:39:12.519793 18451 parallel.cpp:118] Root Solver performance on device 0: 7.164 * 13 = 93.13 img/sec (50000 itr in 6979 sec)
  6541. I0818 15:39:12.519840 18451 parallel.cpp:128] Solver performance on device 1: 7.164 * 13 = 93.13 img/sec (50000 itr in 6979 sec)
  6542. I0818 15:39:12.519870 18451 parallel.cpp:128] Solver performance on device 2: 7.164 * 13 = 93.13 img/sec (50000 itr in 6979 sec)
  6543. I0818 15:39:12.519894 18451 parallel.cpp:128] Solver performance on device 3: 7.164 * 13 = 93.13 img/sec (50000 itr in 6979 sec)
  6544. I0818 15:39:12.519915 18451 parallel.cpp:128] Solver performance on device 4: 7.164 * 13 = 93.13 img/sec (50000 itr in 6979 sec)
  6545. I0818 15:39:12.519943 18451 parallel.cpp:128] Solver performance on device 5: 7.164 * 13 = 93.13 img/sec (50000 itr in 6979 sec)
  6546. I0818 15:39:12.519958 18451 parallel.cpp:128] Solver performance on device 6: 7.164 * 13 = 93.13 img/sec (50000 itr in 6979 sec)
  6547. I0818 15:39:12.519973 18451 parallel.cpp:128] Solver performance on device 7: 7.164 * 13 = 93.13 img/sec (50000 itr in 6979 sec)
  6548. I0818 15:39:12.519981 18451 parallel.cpp:143] Overall multi-GPU performance: 745.073 img/sec
  6549. I0818 15:39:14.346557 18451 caffe.cpp:269] Optimization Done in 2h 4m 8s
  6550. Entering ReduceAndUpdate thread ........................
  6551. Entering ReduceAndUpdate thread ........................
  6552. Entering ReduceAndUpdate thread ........................
  6553. Entering ReduceAndUpdate thread ........................
  6554. Entering ReduceAndUpdate thread ........................
  6555. Entering ReduceAndUpdate thread ........................
  6556. Entering ReduceAndUpdate thread ........................
  6557. Entering ReduceAndUpdate thread ........................
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!