API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Nov 16th, 2021
45
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 13.22 KB | None | 0 0
raw download clone embed print report
  1. 1/14 fused_nn_conv2d_multiply_expand_dims_add_nn_relu_negative_nn_relu_multiply_add_2:1
  2. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  3. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  4. Layout :2
  5. ["f5fce0ea43fa89dfd53142957b424aa7", [1, 512, 512, 20], [1, 1, 20, 40], [40], [40], [1, 1, 1, 40], [1, 512, 512, 40]]
  6. placeholder = PLACEHOLDER [1, 512, 512, 20]
  7. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  8. placeholder = PLACEHOLDER [1, 1, 20, 40]
  9. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  10. placeholder = PLACEHOLDER [40]
  11. placeholder = PLACEHOLDER [40]
  12. T_multiply(ax0) = (placeholder[ax0]*placeholder[ax0])
  13. T_expand_dims(ax0, ax1, ax2, ax3) = T_multiply[ax3]
  14. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + T_expand_dims[ax0, 0, 0, ax3])
  15. T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
  16. placeholder = PLACEHOLDER [1, 1, 1, 40]
  17. T_negative(ax0, ax1, ax2, ax3) = (0f - T_add[ax0, ax1, ax2, ax3])
  18. T_relu(ax0, ax1, ax2, ax3) = max(T_negative[ax0, ax1, ax2, ax3], 0f)
  19. T_multiply(ax0, ax1, ax2, ax3) = (placeholder[ax0, 0, 0, ax3]*T_relu[ax0, ax1, ax2, ax3])
  20. T_add(ax0, ax1, ax2, ax3) = (T_relu[ax0, ax1, ax2, ax3] + T_multiply[ax0, ax1, ax2, ax3])
  21.  
  22. 2/14 fused_nn_conv2d_add_fast_tanh_2:1
  23. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  24. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  25. Layout :2
  26. ["f96d216db5f296dab654242dc885f498", [1, 1, 1, 8], [1, 1, 8, 4], [1, 1, 1, 4], [1, 1, 1, 4]]
  27. placeholder = PLACEHOLDER [1, 1, 1, 8]
  28. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  29. placeholder = PLACEHOLDER [1, 1, 8, 4]
  30. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  31. placeholder = PLACEHOLDER [1, 1, 1, 4]
  32. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + placeholder[ax0, ax1, ax2, ax3])
  33. T_minimum(ax0, ax1, ax2, ax3) = min(T_add[ax0, ax1, ax2, ax3], 9f)
  34. T_maximum(ax0, ax1, ax2, ax3) = max(T_minimum[ax0, ax1, ax2, ax3], -9f)
  35. T_fast_tanh(ax0, ax1, ax2, ax3) = ((T_maximum[ax0, ax1, ax2, ax3]*(((T_maximum[ax0, ax1, ax2, ax3]*T_maximum[ax0, ax1, ax2, ax3] ..(OMITTED).. )*(((T_maximum[ax0, ax1, ax2, ax3]*T_maximum[ax0, ax1, ax2, ax3])*1.19826e-06f) + 0.000118535f)) + 0.00226843f)) + 0.00489353f))
  36.  
  37. 3/14 fused_nn_conv2d_2:8
  38. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  39. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  40. Layout :2
  41. ["4f6690042e1fb8fac22d45ca04ee17a9", [1, 258, 258, 80], [3, 3, 80, 1], [1, 256, 256, 80]]
  42. placeholder = PLACEHOLDER [1, 258, 258, 80]
  43. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  44. placeholder = PLACEHOLDER [3, 3, 80, 1]
  45. DepthwiseConv2d(b, i, j, c) += (PaddedInput[b, (i + di), (j + dj), c]*placeholder[di, dj, c, 0])
  46.  
  47. 4/14 fused_nn_conv2d:2
  48. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  49. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  50. Layout :2
  51. ["4f6690042e1fb8fac22d45ca04ee17a9", [1, 514, 514, 40], [3, 3, 40, 1], [1, 512, 512, 40]]
  52. placeholder = PLACEHOLDER [1, 514, 514, 40]
  53. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  54. placeholder = PLACEHOLDER [3, 3, 40, 1]
  55. DepthwiseConv2d(b, i, j, c) += (PaddedInput[b, (i + di), (j + dj), c]*placeholder[di, dj, c, 0])
  56.  
  57. 5/14 fused_nn_conv2d_multiply_expand_dims_add:1
  58. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  59. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  60. Layout :2
  61. ["a69b4f95edfd32f7f4d07ab68a8a0aa6", [1, 512, 512, 40], [1, 1, 40, 40], [40], [40], [1, 512, 512, 40]]
  62. placeholder = PLACEHOLDER [1, 512, 512, 40]
  63. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  64. placeholder = PLACEHOLDER [1, 1, 40, 40]
  65. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  66. placeholder = PLACEHOLDER [40]
  67. placeholder = PLACEHOLDER [40]
  68. T_multiply(ax0) = (placeholder[ax0]*placeholder[ax0])
  69. T_expand_dims(ax0, ax1, ax2, ax3) = T_multiply[ax3]
  70. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + T_expand_dims[ax0, 0, 0, ax3])
  71.  
  72. 6/14 fused_nn_conv2d_add_fast_tanh:3
  73. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  74. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  75. Layout :2
  76. ["f96d216db5f296dab654242dc885f498", [1, 1, 1, 8], [1, 1, 8, 40], [1, 1, 1, 40], [1, 1, 1, 40]]
  77. placeholder = PLACEHOLDER [1, 1, 1, 8]
  78. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  79. placeholder = PLACEHOLDER [1, 1, 8, 40]
  80. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  81. placeholder = PLACEHOLDER [1, 1, 1, 40]
  82. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + placeholder[ax0, ax1, ax2, ax3])
  83. T_minimum(ax0, ax1, ax2, ax3) = min(T_add[ax0, ax1, ax2, ax3], 9f)
  84. T_maximum(ax0, ax1, ax2, ax3) = max(T_minimum[ax0, ax1, ax2, ax3], -9f)
  85. T_fast_tanh(ax0, ax1, ax2, ax3) = ((T_maximum[ax0, ax1, ax2, ax3]*(((T_maximum[ax0, ax1, ax2, ax3]*T_maximum[ax0, ax1, ax2, ax3] ..(OMITTED).. )*(((T_maximum[ax0, ax1, ax2, ax3]*T_maximum[ax0, ax1, ax2, ax3])*1.19826e-06f) + 0.000118535f)) + 0.00226843f)) + 0.00489353f))
  86.  
  87. 7/14 fused_nn_conv2d_add_fast_tanh_1:25
  88. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  89. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  90. Layout :2
  91. ["f96d216db5f296dab654242dc885f498", [1, 1, 1, 8], [1, 1, 8, 80], [1, 1, 1, 80], [1, 1, 1, 80]]
  92. placeholder = PLACEHOLDER [1, 1, 1, 8]
  93. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  94. placeholder = PLACEHOLDER [1, 1, 8, 80]
  95. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  96. placeholder = PLACEHOLDER [1, 1, 1, 80]
  97. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + placeholder[ax0, ax1, ax2, ax3])
  98. T_minimum(ax0, ax1, ax2, ax3) = min(T_add[ax0, ax1, ax2, ax3], 9f)
  99. T_maximum(ax0, ax1, ax2, ax3) = max(T_minimum[ax0, ax1, ax2, ax3], -9f)
  100. T_fast_tanh(ax0, ax1, ax2, ax3) = ((T_maximum[ax0, ax1, ax2, ax3]*(((T_maximum[ax0, ax1, ax2, ax3]*T_maximum[ax0, ax1, ax2, ax3] ..(OMITTED).. )*(((T_maximum[ax0, ax1, ax2, ax3]*T_maximum[ax0, ax1, ax2, ax3])*1.19826e-06f) + 0.000118535f)) + 0.00226843f)) + 0.00489353f))
  101.  
  102. 8/14 fused_nn_conv2d_multiply_expand_dims_add_nn_relu_negative_nn_relu_multiply_add:1
  103. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  104. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  105. Layout :2
  106. ["f5fce0ea43fa89dfd53142957b424aa7", [1, 512, 512, 4], [1, 1, 4, 40], [40], [40], [1, 1, 1, 40], [1, 512, 512, 40]]
  107. placeholder = PLACEHOLDER [1, 512, 512, 4]
  108. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  109. placeholder = PLACEHOLDER [1, 1, 4, 40]
  110. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  111. placeholder = PLACEHOLDER [40]
  112. placeholder = PLACEHOLDER [40]
  113. T_multiply(ax0) = (placeholder[ax0]*placeholder[ax0])
  114. T_expand_dims(ax0, ax1, ax2, ax3) = T_multiply[ax3]
  115. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + T_expand_dims[ax0, 0, 0, ax3])
  116. T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
  117. placeholder = PLACEHOLDER [1, 1, 1, 40]
  118. T_negative(ax0, ax1, ax2, ax3) = (0f - T_add[ax0, ax1, ax2, ax3])
  119. T_relu(ax0, ax1, ax2, ax3) = max(T_negative[ax0, ax1, ax2, ax3], 0f)
  120. T_multiply(ax0, ax1, ax2, ax3) = (placeholder[ax0, 0, 0, ax3]*T_relu[ax0, ax1, ax2, ax3])
  121. T_add(ax0, ax1, ax2, ax3) = (T_relu[ax0, ax1, ax2, ax3] + T_multiply[ax0, ax1, ax2, ax3])
  122.  
  123. 9/14 fused_nn_conv2d_multiply_expand_dims_add_add:1
  124. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  125. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  126. Layout :2
  127. ["3b642a36ecdff080fb7797fdf88a3592", [1, 512, 512, 40], [1, 1, 40, 4], [4], [4], [1, 512, 512, 4], [1, 512, 512, 4]]
  128. placeholder = PLACEHOLDER [1, 512, 512, 40]
  129. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  130. placeholder = PLACEHOLDER [1, 1, 40, 4]
  131. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  132. placeholder = PLACEHOLDER [4]
  133. placeholder = PLACEHOLDER [4]
  134. T_multiply(ax0) = (placeholder[ax0]*placeholder[ax0])
  135. T_expand_dims(ax0, ax1, ax2, ax3) = T_multiply[ax3]
  136. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + T_expand_dims[ax0, 0, 0, ax3])
  137. placeholder = PLACEHOLDER [1, 512, 512, 4]
  138. T_add(ax0, ax1, ax2, ax3) = (T_add[ax0, ax1, ax2, ax3] + placeholder[ax0, ax1, ax2, ax3])
  139.  
  140. 10/14 fused_nn_conv2d_1:1
  141. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  142. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  143. Layout :2
  144. ["4f6690042e1fb8fac22d45ca04ee17a9", [1, 258, 258, 160], [3, 3, 160, 1], [1, 256, 256, 160]]
  145. placeholder = PLACEHOLDER [1, 258, 258, 160]
  146. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  147. placeholder = PLACEHOLDER [3, 3, 160, 1]
  148. DepthwiseConv2d(b, i, j, c) += (PaddedInput[b, (i + di), (j + dj), c]*placeholder[di, dj, c, 0])
  149.  
  150. 11/14 fused_nn_conv2d_add:1
  151. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  152. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  153. Layout :2
  154. ["1cc666833c122282e3fcf3595901b12b", [1, 1, 1, 1], [1, 1, 1, 232], [1, 1, 1, 232], [1, 1, 1, 232]]
  155. placeholder = PLACEHOLDER [1, 1, 1, 1]
  156. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  157. placeholder = PLACEHOLDER [1, 1, 1, 232]
  158. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  159. placeholder = PLACEHOLDER [1, 1, 1, 232]
  160. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + placeholder[ax0, ax1, ax2, ax3])
  161.  
  162. 12/14 fused_nn_conv2d_multiply_expand_dims_add_multiply_add:8
  163. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  164. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  165. Layout :2
  166. ["bb9b318362b5e0751c1f053b058762f7", [1, 256, 256, 80], [1, 1, 80, 80], [80], [80], [1, 256, 256, 80], [1, 256, 256, 80]]
  167. placeholder = PLACEHOLDER [1, 256, 256, 80]
  168. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  169. placeholder = PLACEHOLDER [1, 1, 80, 80]
  170. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  171. placeholder = PLACEHOLDER [80]
  172. placeholder = PLACEHOLDER [80]
  173. T_multiply(ax0) = (placeholder[ax0]*placeholder[ax0])
  174. T_expand_dims(ax0, ax1, ax2, ax3) = T_multiply[ax3]
  175. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + T_expand_dims[ax0, 0, 0, ax3])
  176. compile_engine_const() = 1.5f
  177. placeholder = PLACEHOLDER [1, 256, 256, 80]
  178. T_multiply(ax0, ax1, ax2, ax3) = (compile_engine_const[]*placeholder[ax0, ax1, ax2, ax3])
  179. T_add(ax0, ax1, ax2, ax3) = (T_add[ax0, ax1, ax2, ax3] + T_multiply[ax0, ax1, ax2, ax3])
  180.  
  181. 13/14 fused_nn_conv2d_multiply_expand_dims_add_1:1
  182. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  183. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  184. Layout :2
  185. ["a69b4f95edfd32f7f4d07ab68a8a0aa6", [1, 256, 256, 160], [1, 1, 160, 80], [80], [80], [1, 256, 256, 80]]
  186. placeholder = PLACEHOLDER [1, 256, 256, 160]
  187. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  188. placeholder = PLACEHOLDER [1, 1, 160, 80]
  189. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  190. placeholder = PLACEHOLDER [80]
  191. placeholder = PLACEHOLDER [80]
  192. T_multiply(ax0) = (placeholder[ax0]*placeholder[ax0])
  193. T_expand_dims(ax0, ax1, ax2, ax3) = T_multiply[ax3]
  194. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + T_expand_dims[ax0, 0, 0, ax3])
  195.  
  196. 14/14 fused_nn_conv2d_multiply_expand_dims_add_nn_relu_negative_nn_relu_multiply_add_1:16
  197. Target :opencl -keys=mali,opencl,gpu -device=mali -max_num_threads=256 -thread_warp_size=1
  198. TargetHost:llvm -keys=cpu -link-params=0 -mtriple=arm64-linux-android
  199. Layout :2
  200. ["f5fce0ea43fa89dfd53142957b424aa7", [1, 256, 256, 80], [1, 1, 80, 80], [80], [80], [1, 1, 1, 80], [1, 256, 256, 80]]
  201. placeholder = PLACEHOLDER [1, 256, 256, 80]
  202. PaddedInput(i0, i1, i2, i3) = placeholder[i0, i1, i2, i3]
  203. placeholder = PLACEHOLDER [1, 1, 80, 80]
  204. Conv2dOutput(nn, yy, xx, ff) += (PaddedInput[nn, (yy + ry), (xx + rx), rc]*placeholder[ry, rx, rc, ff])
  205. placeholder = PLACEHOLDER [80]
  206. placeholder = PLACEHOLDER [80]
  207. T_multiply(ax0) = (placeholder[ax0]*placeholder[ax0])
  208. T_expand_dims(ax0, ax1, ax2, ax3) = T_multiply[ax3]
  209. T_add(ax0, ax1, ax2, ax3) = (Conv2dOutput[ax0, ax1, ax2, ax3] + T_expand_dims[ax0, 0, 0, ax3])
  210. T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
  211. placeholder = PLACEHOLDER [1, 1, 1, 80]
  212. T_negative(ax0, ax1, ax2, ax3) = (0f - T_add[ax0, ax1, ax2, ax3])
  213. T_relu(ax0, ax1, ax2, ax3) = max(T_negative[ax0, ax1, ax2, ax3], 0f)
  214. T_multiply(ax0, ax1, ax2, ax3) = (placeholder[ax0, 0, 0, ax3]*T_relu[ax0, ax1, ax2, ax3])
  215. T_add(ax0, ax1, ax2, ax3) = (T_relu[ax0, ax1, ax2, ax3] + T_multiply[ax0, ax1, ax2, ax3])
  216.  
  217.  
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!