Guest User

yolov11_qat

a guest
Apr 9th, 2025
904
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Python 155.40 KB | None | 0 0
  1. DetectionModel(
  2.   (model): Sequential(
  3.     (0): Conv(
  4.       (quant): QuantStub(
  5.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  6.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  7.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  8.         )
  9.       )
  10.       (dequant): DeQuantStub()
  11.       (conv): ConvReLU2d(
  12.         3, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
  13.         (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  14.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  15.           (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  16.         )
  17.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  18.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  19.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  20.         )
  21.       )
  22.       (bn): Identity()
  23.       (act): Identity()
  24.     )
  25.     (1): Conv(
  26.       (quant): QuantStub(
  27.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  28.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  29.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  30.         )
  31.       )
  32.       (dequant): DeQuantStub()
  33.       (conv): ConvReLU2d(
  34.         16, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
  35.         (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  36.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  37.           (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  38.         )
  39.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  40.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  41.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  42.         )
  43.       )
  44.       (bn): Identity()
  45.       (act): Identity()
  46.     )
  47.     (2): C3k2(
  48.       (cv1): Conv(
  49.         (quant): QuantStub(
  50.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  51.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  52.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  53.           )
  54.         )
  55.         (dequant): DeQuantStub()
  56.         (conv): ConvReLU2d(
  57.           32, 32, kernel_size=(1, 1), stride=(1, 1)
  58.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  59.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  60.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  61.           )
  62.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  63.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  64.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  65.           )
  66.         )
  67.         (bn): Identity()
  68.         (act): Identity()
  69.       )
  70.       (cv2): Conv(
  71.         (quant): QuantStub(
  72.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  73.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  74.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  75.           )
  76.         )
  77.         (dequant): DeQuantStub()
  78.         (conv): ConvReLU2d(
  79.           48, 64, kernel_size=(1, 1), stride=(1, 1)
  80.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  81.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  82.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  83.           )
  84.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  85.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  86.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  87.           )
  88.         )
  89.         (bn): Identity()
  90.         (act): Identity()
  91.       )
  92.       (m): ModuleList(
  93.         (0): Bottleneck(
  94.           (cv1): Conv(
  95.             (quant): QuantStub(
  96.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  97.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  98.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  99.               )
  100.             )
  101.             (dequant): DeQuantStub()
  102.             (conv): ConvReLU2d(
  103.               16, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  104.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  105.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  106.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  107.               )
  108.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  109.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  110.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  111.               )
  112.             )
  113.             (bn): Identity()
  114.             (act): Identity()
  115.           )
  116.           (cv2): Conv(
  117.             (quant): QuantStub(
  118.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  119.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  120.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  121.               )
  122.             )
  123.             (dequant): DeQuantStub()
  124.             (conv): ConvReLU2d(
  125.               8, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  126.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  127.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  128.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  129.               )
  130.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  131.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  132.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  133.               )
  134.             )
  135.             (bn): Identity()
  136.             (act): Identity()
  137.           )
  138.         )
  139.       )
  140.     )
  141.     (3): Conv(
  142.       (quant): QuantStub(
  143.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  144.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  145.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  146.         )
  147.       )
  148.       (dequant): DeQuantStub()
  149.       (conv): ConvReLU2d(
  150.         64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
  151.         (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  152.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  153.           (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  154.         )
  155.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  156.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  157.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  158.         )
  159.       )
  160.       (bn): Identity()
  161.       (act): Identity()
  162.     )
  163.     (4): C3k2(
  164.       (cv1): Conv(
  165.         (quant): QuantStub(
  166.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  167.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  168.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  169.           )
  170.         )
  171.         (dequant): DeQuantStub()
  172.         (conv): ConvReLU2d(
  173.           64, 64, kernel_size=(1, 1), stride=(1, 1)
  174.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  175.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  176.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  177.           )
  178.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  179.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  180.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  181.           )
  182.         )
  183.         (bn): Identity()
  184.         (act): Identity()
  185.       )
  186.       (cv2): Conv(
  187.         (quant): QuantStub(
  188.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  189.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  190.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  191.           )
  192.         )
  193.         (dequant): DeQuantStub()
  194.         (conv): ConvReLU2d(
  195.           96, 128, kernel_size=(1, 1), stride=(1, 1)
  196.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  197.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  198.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  199.           )
  200.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  201.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  202.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  203.           )
  204.         )
  205.         (bn): Identity()
  206.         (act): Identity()
  207.       )
  208.       (m): ModuleList(
  209.         (0): Bottleneck(
  210.           (cv1): Conv(
  211.             (quant): QuantStub(
  212.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  213.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  214.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  215.               )
  216.             )
  217.             (dequant): DeQuantStub()
  218.             (conv): ConvReLU2d(
  219.               32, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  220.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  221.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  222.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  223.               )
  224.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  225.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  226.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  227.               )
  228.             )
  229.             (bn): Identity()
  230.             (act): Identity()
  231.           )
  232.           (cv2): Conv(
  233.             (quant): QuantStub(
  234.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  235.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  236.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  237.               )
  238.             )
  239.             (dequant): DeQuantStub()
  240.             (conv): ConvReLU2d(
  241.               16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  242.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  243.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  244.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  245.               )
  246.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  247.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  248.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  249.               )
  250.             )
  251.             (bn): Identity()
  252.             (act): Identity()
  253.           )
  254.         )
  255.       )
  256.     )
  257.     (5): Conv(
  258.       (quant): QuantStub(
  259.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  260.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  261.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  262.         )
  263.       )
  264.       (dequant): DeQuantStub()
  265.       (conv): ConvReLU2d(
  266.         128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
  267.         (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  268.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  269.           (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  270.         )
  271.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  272.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  273.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  274.         )
  275.       )
  276.       (bn): Identity()
  277.       (act): Identity()
  278.     )
  279.     (6): C3k2(
  280.       (cv1): Conv(
  281.         (quant): QuantStub(
  282.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  283.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  284.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  285.           )
  286.         )
  287.         (dequant): DeQuantStub()
  288.         (conv): ConvReLU2d(
  289.           128, 128, kernel_size=(1, 1), stride=(1, 1)
  290.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  291.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  292.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  293.           )
  294.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  295.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  296.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  297.           )
  298.         )
  299.         (bn): Identity()
  300.         (act): Identity()
  301.       )
  302.       (cv2): Conv(
  303.         (quant): QuantStub(
  304.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  305.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  306.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  307.           )
  308.         )
  309.         (dequant): DeQuantStub()
  310.         (conv): ConvReLU2d(
  311.           192, 128, kernel_size=(1, 1), stride=(1, 1)
  312.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  313.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  314.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  315.           )
  316.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  317.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  318.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  319.           )
  320.         )
  321.         (bn): Identity()
  322.         (act): Identity()
  323.       )
  324.       (m): ModuleList(
  325.         (0): C3k(
  326.           (cv1): Conv(
  327.             (quant): QuantStub(
  328.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  329.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  330.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  331.               )
  332.             )
  333.             (dequant): DeQuantStub()
  334.             (conv): ConvReLU2d(
  335.               64, 32, kernel_size=(1, 1), stride=(1, 1)
  336.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  337.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  338.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  339.               )
  340.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  341.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  342.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  343.               )
  344.             )
  345.             (bn): Identity()
  346.             (act): Identity()
  347.           )
  348.           (cv2): Conv(
  349.             (quant): QuantStub(
  350.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  351.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  352.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  353.               )
  354.             )
  355.             (dequant): DeQuantStub()
  356.             (conv): ConvReLU2d(
  357.               64, 32, kernel_size=(1, 1), stride=(1, 1)
  358.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  359.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  360.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  361.               )
  362.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  363.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  364.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  365.               )
  366.             )
  367.             (bn): Identity()
  368.             (act): Identity()
  369.           )
  370.           (cv3): Conv(
  371.             (quant): QuantStub(
  372.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  373.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  374.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  375.               )
  376.             )
  377.             (dequant): DeQuantStub()
  378.             (conv): ConvReLU2d(
  379.               64, 64, kernel_size=(1, 1), stride=(1, 1)
  380.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  381.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  382.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  383.               )
  384.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  385.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  386.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  387.               )
  388.             )
  389.             (bn): Identity()
  390.             (act): Identity()
  391.           )
  392.           (m): Sequential(
  393.             (0): Bottleneck(
  394.               (cv1): Conv(
  395.                 (quant): QuantStub(
  396.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  397.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  398.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  399.                   )
  400.                 )
  401.                 (dequant): DeQuantStub()
  402.                 (conv): ConvReLU2d(
  403.                   32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  404.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  405.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  406.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  407.                   )
  408.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  409.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  410.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  411.                   )
  412.                 )
  413.                 (bn): Identity()
  414.                 (act): Identity()
  415.               )
  416.               (cv2): Conv(
  417.                 (quant): QuantStub(
  418.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  419.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  420.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  421.                   )
  422.                 )
  423.                 (dequant): DeQuantStub()
  424.                 (conv): ConvReLU2d(
  425.                   32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  426.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  427.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  428.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  429.                   )
  430.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  431.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  432.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  433.                   )
  434.                 )
  435.                 (bn): Identity()
  436.                 (act): Identity()
  437.               )
  438.             )
  439.             (1): Bottleneck(
  440.               (cv1): Conv(
  441.                 (quant): QuantStub(
  442.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  443.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  444.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  445.                   )
  446.                 )
  447.                 (dequant): DeQuantStub()
  448.                 (conv): ConvReLU2d(
  449.                   32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  450.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  451.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  452.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  453.                   )
  454.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  455.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  456.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  457.                   )
  458.                 )
  459.                 (bn): Identity()
  460.                 (act): Identity()
  461.               )
  462.               (cv2): Conv(
  463.                 (quant): QuantStub(
  464.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  465.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  466.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  467.                   )
  468.                 )
  469.                 (dequant): DeQuantStub()
  470.                 (conv): ConvReLU2d(
  471.                   32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  472.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  473.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  474.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  475.                   )
  476.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  477.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  478.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  479.                   )
  480.                 )
  481.                 (bn): Identity()
  482.                 (act): Identity()
  483.               )
  484.             )
  485.           )
  486.         )
  487.       )
  488.     )
  489.     (7): Conv(
  490.       (quant): QuantStub(
  491.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  492.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  493.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  494.         )
  495.       )
  496.       (dequant): DeQuantStub()
  497.       (conv): ConvReLU2d(
  498.         128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
  499.         (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  500.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  501.           (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  502.         )
  503.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  504.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  505.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  506.         )
  507.       )
  508.       (bn): Identity()
  509.       (act): Identity()
  510.     )
  511.     (8): C3k2(
  512.       (cv1): Conv(
  513.         (quant): QuantStub(
  514.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  515.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  516.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  517.           )
  518.         )
  519.         (dequant): DeQuantStub()
  520.         (conv): ConvReLU2d(
  521.           256, 256, kernel_size=(1, 1), stride=(1, 1)
  522.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  523.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  524.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  525.           )
  526.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  527.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  528.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  529.           )
  530.         )
  531.         (bn): Identity()
  532.         (act): Identity()
  533.       )
  534.       (cv2): Conv(
  535.         (quant): QuantStub(
  536.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  537.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  538.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  539.           )
  540.         )
  541.         (dequant): DeQuantStub()
  542.         (conv): ConvReLU2d(
  543.           384, 256, kernel_size=(1, 1), stride=(1, 1)
  544.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  545.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  546.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  547.           )
  548.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  549.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  550.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  551.           )
  552.         )
  553.         (bn): Identity()
  554.         (act): Identity()
  555.       )
  556.       (m): ModuleList(
  557.         (0): C3k(
  558.           (cv1): Conv(
  559.             (quant): QuantStub(
  560.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  561.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  562.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  563.               )
  564.             )
  565.             (dequant): DeQuantStub()
  566.             (conv): ConvReLU2d(
  567.               128, 64, kernel_size=(1, 1), stride=(1, 1)
  568.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  569.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  570.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  571.               )
  572.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  573.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  574.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  575.               )
  576.             )
  577.             (bn): Identity()
  578.             (act): Identity()
  579.           )
  580.           (cv2): Conv(
  581.             (quant): QuantStub(
  582.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  583.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  584.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  585.               )
  586.             )
  587.             (dequant): DeQuantStub()
  588.             (conv): ConvReLU2d(
  589.               128, 64, kernel_size=(1, 1), stride=(1, 1)
  590.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  591.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  592.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  593.               )
  594.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  595.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  596.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  597.               )
  598.             )
  599.             (bn): Identity()
  600.             (act): Identity()
  601.           )
  602.           (cv3): Conv(
  603.             (quant): QuantStub(
  604.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  605.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  606.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  607.               )
  608.             )
  609.             (dequant): DeQuantStub()
  610.             (conv): ConvReLU2d(
  611.               128, 128, kernel_size=(1, 1), stride=(1, 1)
  612.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  613.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  614.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  615.               )
  616.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  617.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  618.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  619.               )
  620.             )
  621.             (bn): Identity()
  622.             (act): Identity()
  623.           )
  624.           (m): Sequential(
  625.             (0): Bottleneck(
  626.               (cv1): Conv(
  627.                 (quant): QuantStub(
  628.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  629.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  630.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  631.                   )
  632.                 )
  633.                 (dequant): DeQuantStub()
  634.                 (conv): ConvReLU2d(
  635.                   64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  636.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  637.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  638.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  639.                   )
  640.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  641.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  642.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  643.                   )
  644.                 )
  645.                 (bn): Identity()
  646.                 (act): Identity()
  647.               )
  648.               (cv2): Conv(
  649.                 (quant): QuantStub(
  650.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  651.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  652.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  653.                   )
  654.                 )
  655.                 (dequant): DeQuantStub()
  656.                 (conv): ConvReLU2d(
  657.                   64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  658.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  659.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  660.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  661.                   )
  662.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  663.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  664.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  665.                   )
  666.                 )
  667.                 (bn): Identity()
  668.                 (act): Identity()
  669.               )
  670.             )
  671.             (1): Bottleneck(
  672.               (cv1): Conv(
  673.                 (quant): QuantStub(
  674.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  675.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  676.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  677.                   )
  678.                 )
  679.                 (dequant): DeQuantStub()
  680.                 (conv): ConvReLU2d(
  681.                   64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  682.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  683.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  684.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  685.                   )
  686.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  687.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  688.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  689.                   )
  690.                 )
  691.                 (bn): Identity()
  692.                 (act): Identity()
  693.               )
  694.               (cv2): Conv(
  695.                 (quant): QuantStub(
  696.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  697.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  698.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  699.                   )
  700.                 )
  701.                 (dequant): DeQuantStub()
  702.                 (conv): ConvReLU2d(
  703.                   64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  704.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  705.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  706.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  707.                   )
  708.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  709.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  710.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  711.                   )
  712.                 )
  713.                 (bn): Identity()
  714.                 (act): Identity()
  715.               )
  716.             )
  717.           )
  718.         )
  719.       )
  720.     )
  721.     (9): SPPF(
  722.       (cv1): Conv(
  723.         (quant): QuantStub(
  724.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  725.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  726.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  727.           )
  728.         )
  729.         (dequant): DeQuantStub()
  730.         (conv): ConvReLU2d(
  731.           256, 128, kernel_size=(1, 1), stride=(1, 1)
  732.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  733.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  734.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  735.           )
  736.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  737.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  738.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  739.           )
  740.         )
  741.         (bn): Identity()
  742.         (act): Identity()
  743.       )
  744.       (cv2): Conv(
  745.         (quant): QuantStub(
  746.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  747.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  748.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  749.           )
  750.         )
  751.         (dequant): DeQuantStub()
  752.         (conv): ConvReLU2d(
  753.           512, 256, kernel_size=(1, 1), stride=(1, 1)
  754.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  755.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  756.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  757.           )
  758.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  759.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  760.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  761.           )
  762.         )
  763.         (bn): Identity()
  764.         (act): Identity()
  765.       )
  766.       (m): MaxPool2d(kernel_size=5, stride=1, padding=2, dilation=1, ceil_mode=False)
  767.     )
  768.     (10): C2PSA(
  769.       (cv1): Conv(
  770.         (quant): QuantStub(
  771.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  772.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  773.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  774.           )
  775.         )
  776.         (dequant): DeQuantStub()
  777.         (conv): ConvReLU2d(
  778.           256, 256, kernel_size=(1, 1), stride=(1, 1)
  779.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  780.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  781.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  782.           )
  783.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  784.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  785.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  786.           )
  787.         )
  788.         (bn): Identity()
  789.         (act): Identity()
  790.       )
  791.       (cv2): Conv(
  792.         (quant): QuantStub(
  793.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  794.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  795.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  796.           )
  797.         )
  798.         (dequant): DeQuantStub()
  799.         (conv): ConvReLU2d(
  800.           256, 256, kernel_size=(1, 1), stride=(1, 1)
  801.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  802.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  803.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  804.           )
  805.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  806.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  807.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  808.           )
  809.         )
  810.         (bn): Identity()
  811.         (act): Identity()
  812.       )
  813.       (m): Sequential(
  814.         (0): PSABlock(
  815.           (attn): Attention(
  816.             (qkv): Conv(
  817.               (quant): QuantStub(
  818.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  819.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  820.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  821.                 )
  822.               )
  823.               (dequant): DeQuantStub()
  824.               (conv): Conv2d(
  825.                 128, 256, kernel_size=(1, 1), stride=(1, 1)
  826.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  827.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  828.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  829.                 )
  830.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  831.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  832.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  833.                 )
  834.               )
  835.               (bn): Identity()
  836.               (act): Identity()
  837.             )
  838.             (proj): Conv(
  839.               (quant): QuantStub(
  840.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  841.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  842.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  843.                 )
  844.               )
  845.               (dequant): DeQuantStub()
  846.               (conv): Conv2d(
  847.                 128, 128, kernel_size=(1, 1), stride=(1, 1)
  848.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  849.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  850.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  851.                 )
  852.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  853.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  854.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  855.                 )
  856.               )
  857.               (bn): Identity()
  858.               (act): Identity()
  859.             )
  860.             (pe): Conv(
  861.               (quant): QuantStub(
  862.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  863.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  864.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  865.                 )
  866.               )
  867.               (dequant): DeQuantStub()
  868.               (conv): Conv2d(
  869.                 128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128
  870.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  871.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  872.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  873.                 )
  874.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  875.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  876.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  877.                 )
  878.               )
  879.               (bn): Identity()
  880.               (act): Identity()
  881.             )
  882.           )
  883.           (ffn): Sequential(
  884.             (0): Conv(
  885.               (quant): QuantStub(
  886.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  887.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  888.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  889.                 )
  890.               )
  891.               (dequant): DeQuantStub()
  892.               (conv): ConvReLU2d(
  893.                 128, 256, kernel_size=(1, 1), stride=(1, 1)
  894.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  895.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  896.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  897.                 )
  898.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  899.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  900.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  901.                 )
  902.               )
  903.               (bn): Identity()
  904.               (act): Identity()
  905.             )
  906.             (1): Conv(
  907.               (quant): QuantStub(
  908.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  909.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  910.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  911.                 )
  912.               )
  913.               (dequant): DeQuantStub()
  914.               (conv): Conv2d(
  915.                 256, 128, kernel_size=(1, 1), stride=(1, 1)
  916.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  917.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  918.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  919.                 )
  920.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  921.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  922.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  923.                 )
  924.               )
  925.               (bn): Identity()
  926.               (act): Identity()
  927.             )
  928.           )
  929.         )
  930.       )
  931.     )
  932.     (11): Upsample(scale_factor=2.0, mode='nearest')
  933.     (12): Concat()
  934.     (13): C3k2(
  935.       (cv1): Conv(
  936.         (quant): QuantStub(
  937.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  938.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  939.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  940.           )
  941.         )
  942.         (dequant): DeQuantStub()
  943.         (conv): ConvReLU2d(
  944.           384, 128, kernel_size=(1, 1), stride=(1, 1)
  945.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  946.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  947.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  948.           )
  949.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  950.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  951.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  952.           )
  953.         )
  954.         (bn): Identity()
  955.         (act): Identity()
  956.       )
  957.       (cv2): Conv(
  958.         (quant): QuantStub(
  959.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  960.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  961.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  962.           )
  963.         )
  964.         (dequant): DeQuantStub()
  965.         (conv): ConvReLU2d(
  966.           192, 128, kernel_size=(1, 1), stride=(1, 1)
  967.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  968.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  969.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  970.           )
  971.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  972.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  973.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  974.           )
  975.         )
  976.         (bn): Identity()
  977.         (act): Identity()
  978.       )
  979.       (m): ModuleList(
  980.         (0): Bottleneck(
  981.           (cv1): Conv(
  982.             (quant): QuantStub(
  983.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  984.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  985.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  986.               )
  987.             )
  988.             (dequant): DeQuantStub()
  989.             (conv): ConvReLU2d(
  990.               64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  991.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  992.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  993.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  994.               )
  995.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  996.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  997.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  998.               )
  999.             )
  1000.             (bn): Identity()
  1001.             (act): Identity()
  1002.           )
  1003.           (cv2): Conv(
  1004.             (quant): QuantStub(
  1005.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1006.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1007.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1008.               )
  1009.             )
  1010.             (dequant): DeQuantStub()
  1011.             (conv): ConvReLU2d(
  1012.               32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1013.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1014.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1015.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1016.               )
  1017.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1018.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1019.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1020.               )
  1021.             )
  1022.             (bn): Identity()
  1023.             (act): Identity()
  1024.           )
  1025.         )
  1026.       )
  1027.     )
  1028.     (14): Upsample(scale_factor=2.0, mode='nearest')
  1029.     (15): Concat()
  1030.     (16): C3k2(
  1031.       (cv1): Conv(
  1032.         (quant): QuantStub(
  1033.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1034.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1035.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1036.           )
  1037.         )
  1038.         (dequant): DeQuantStub()
  1039.         (conv): ConvReLU2d(
  1040.           256, 64, kernel_size=(1, 1), stride=(1, 1)
  1041.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1042.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1043.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1044.           )
  1045.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1046.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1047.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1048.           )
  1049.         )
  1050.         (bn): Identity()
  1051.         (act): Identity()
  1052.       )
  1053.       (cv2): Conv(
  1054.         (quant): QuantStub(
  1055.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1056.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1057.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1058.           )
  1059.         )
  1060.         (dequant): DeQuantStub()
  1061.         (conv): ConvReLU2d(
  1062.           96, 64, kernel_size=(1, 1), stride=(1, 1)
  1063.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1064.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1065.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1066.           )
  1067.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1068.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1069.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1070.           )
  1071.         )
  1072.         (bn): Identity()
  1073.         (act): Identity()
  1074.       )
  1075.       (m): ModuleList(
  1076.         (0): Bottleneck(
  1077.           (cv1): Conv(
  1078.             (quant): QuantStub(
  1079.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1080.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1081.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1082.               )
  1083.             )
  1084.             (dequant): DeQuantStub()
  1085.             (conv): ConvReLU2d(
  1086.               32, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1087.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1088.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1089.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1090.               )
  1091.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1092.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1093.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1094.               )
  1095.             )
  1096.             (bn): Identity()
  1097.             (act): Identity()
  1098.           )
  1099.           (cv2): Conv(
  1100.             (quant): QuantStub(
  1101.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1102.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1103.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1104.               )
  1105.             )
  1106.             (dequant): DeQuantStub()
  1107.             (conv): ConvReLU2d(
  1108.               16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1109.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1110.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1111.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1112.               )
  1113.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1114.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1115.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1116.               )
  1117.             )
  1118.             (bn): Identity()
  1119.             (act): Identity()
  1120.           )
  1121.         )
  1122.       )
  1123.     )
  1124.     (17): Conv(
  1125.       (quant): QuantStub(
  1126.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1127.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1128.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1129.         )
  1130.       )
  1131.       (dequant): DeQuantStub()
  1132.       (conv): ConvReLU2d(
  1133.         64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
  1134.         (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1135.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1136.           (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1137.         )
  1138.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1139.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1140.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1141.         )
  1142.       )
  1143.       (bn): Identity()
  1144.       (act): Identity()
  1145.     )
  1146.     (18): Concat()
  1147.     (19): C3k2(
  1148.       (cv1): Conv(
  1149.         (quant): QuantStub(
  1150.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1151.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1152.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1153.           )
  1154.         )
  1155.         (dequant): DeQuantStub()
  1156.         (conv): ConvReLU2d(
  1157.           192, 128, kernel_size=(1, 1), stride=(1, 1)
  1158.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1159.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1160.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1161.           )
  1162.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1163.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1164.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1165.           )
  1166.         )
  1167.         (bn): Identity()
  1168.         (act): Identity()
  1169.       )
  1170.       (cv2): Conv(
  1171.         (quant): QuantStub(
  1172.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1173.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1174.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1175.           )
  1176.         )
  1177.         (dequant): DeQuantStub()
  1178.         (conv): ConvReLU2d(
  1179.           192, 128, kernel_size=(1, 1), stride=(1, 1)
  1180.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1181.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1182.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1183.           )
  1184.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1185.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1186.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1187.           )
  1188.         )
  1189.         (bn): Identity()
  1190.         (act): Identity()
  1191.       )
  1192.       (m): ModuleList(
  1193.         (0): Bottleneck(
  1194.           (cv1): Conv(
  1195.             (quant): QuantStub(
  1196.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1197.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1198.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1199.               )
  1200.             )
  1201.             (dequant): DeQuantStub()
  1202.             (conv): ConvReLU2d(
  1203.               64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1204.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1205.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1206.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1207.               )
  1208.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1209.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1210.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1211.               )
  1212.             )
  1213.             (bn): Identity()
  1214.             (act): Identity()
  1215.           )
  1216.           (cv2): Conv(
  1217.             (quant): QuantStub(
  1218.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1219.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1220.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1221.               )
  1222.             )
  1223.             (dequant): DeQuantStub()
  1224.             (conv): ConvReLU2d(
  1225.               32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1226.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1227.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1228.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1229.               )
  1230.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1231.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1232.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1233.               )
  1234.             )
  1235.             (bn): Identity()
  1236.             (act): Identity()
  1237.           )
  1238.         )
  1239.       )
  1240.     )
  1241.     (20): Conv(
  1242.       (quant): QuantStub(
  1243.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1244.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1245.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1246.         )
  1247.       )
  1248.       (dequant): DeQuantStub()
  1249.       (conv): ConvReLU2d(
  1250.         128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
  1251.         (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1252.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1253.           (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1254.         )
  1255.         (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1256.           fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1257.           (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1258.         )
  1259.       )
  1260.       (bn): Identity()
  1261.       (act): Identity()
  1262.     )
  1263.     (21): Concat()
  1264.     (22): C3k2(
  1265.       (cv1): Conv(
  1266.         (quant): QuantStub(
  1267.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1268.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1269.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1270.           )
  1271.         )
  1272.         (dequant): DeQuantStub()
  1273.         (conv): ConvReLU2d(
  1274.           384, 256, kernel_size=(1, 1), stride=(1, 1)
  1275.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1276.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1277.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1278.           )
  1279.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1280.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1281.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1282.           )
  1283.         )
  1284.         (bn): Identity()
  1285.         (act): Identity()
  1286.       )
  1287.       (cv2): Conv(
  1288.         (quant): QuantStub(
  1289.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1290.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1291.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1292.           )
  1293.         )
  1294.         (dequant): DeQuantStub()
  1295.         (conv): ConvReLU2d(
  1296.           384, 256, kernel_size=(1, 1), stride=(1, 1)
  1297.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1298.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1299.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1300.           )
  1301.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1302.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1303.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1304.           )
  1305.         )
  1306.         (bn): Identity()
  1307.         (act): Identity()
  1308.       )
  1309.       (m): ModuleList(
  1310.         (0): C3k(
  1311.           (cv1): Conv(
  1312.             (quant): QuantStub(
  1313.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1314.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1315.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1316.               )
  1317.             )
  1318.             (dequant): DeQuantStub()
  1319.             (conv): ConvReLU2d(
  1320.               128, 64, kernel_size=(1, 1), stride=(1, 1)
  1321.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1322.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1323.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1324.               )
  1325.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1326.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1327.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1328.               )
  1329.             )
  1330.             (bn): Identity()
  1331.             (act): Identity()
  1332.           )
  1333.           (cv2): Conv(
  1334.             (quant): QuantStub(
  1335.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1336.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1337.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1338.               )
  1339.             )
  1340.             (dequant): DeQuantStub()
  1341.             (conv): ConvReLU2d(
  1342.               128, 64, kernel_size=(1, 1), stride=(1, 1)
  1343.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1344.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1345.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1346.               )
  1347.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1348.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1349.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1350.               )
  1351.             )
  1352.             (bn): Identity()
  1353.             (act): Identity()
  1354.           )
  1355.           (cv3): Conv(
  1356.             (quant): QuantStub(
  1357.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1358.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1359.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1360.               )
  1361.             )
  1362.             (dequant): DeQuantStub()
  1363.             (conv): ConvReLU2d(
  1364.               128, 128, kernel_size=(1, 1), stride=(1, 1)
  1365.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1366.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1367.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1368.               )
  1369.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1370.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1371.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1372.               )
  1373.             )
  1374.             (bn): Identity()
  1375.             (act): Identity()
  1376.           )
  1377.           (m): Sequential(
  1378.             (0): Bottleneck(
  1379.               (cv1): Conv(
  1380.                 (quant): QuantStub(
  1381.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1382.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1383.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1384.                   )
  1385.                 )
  1386.                 (dequant): DeQuantStub()
  1387.                 (conv): ConvReLU2d(
  1388.                   64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1389.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1390.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1391.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1392.                   )
  1393.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1394.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1395.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1396.                   )
  1397.                 )
  1398.                 (bn): Identity()
  1399.                 (act): Identity()
  1400.               )
  1401.               (cv2): Conv(
  1402.                 (quant): QuantStub(
  1403.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1404.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1405.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1406.                   )
  1407.                 )
  1408.                 (dequant): DeQuantStub()
  1409.                 (conv): ConvReLU2d(
  1410.                   64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1411.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1412.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1413.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1414.                   )
  1415.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1416.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1417.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1418.                   )
  1419.                 )
  1420.                 (bn): Identity()
  1421.                 (act): Identity()
  1422.               )
  1423.             )
  1424.             (1): Bottleneck(
  1425.               (cv1): Conv(
  1426.                 (quant): QuantStub(
  1427.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1428.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1429.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1430.                   )
  1431.                 )
  1432.                 (dequant): DeQuantStub()
  1433.                 (conv): ConvReLU2d(
  1434.                   64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1435.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1436.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1437.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1438.                   )
  1439.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1440.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1441.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1442.                   )
  1443.                 )
  1444.                 (bn): Identity()
  1445.                 (act): Identity()
  1446.               )
  1447.               (cv2): Conv(
  1448.                 (quant): QuantStub(
  1449.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1450.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1451.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1452.                   )
  1453.                 )
  1454.                 (dequant): DeQuantStub()
  1455.                 (conv): ConvReLU2d(
  1456.                   64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1457.                   (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1458.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1459.                     (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1460.                   )
  1461.                   (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1462.                     fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1463.                     (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1464.                   )
  1465.                 )
  1466.                 (bn): Identity()
  1467.                 (act): Identity()
  1468.               )
  1469.             )
  1470.           )
  1471.         )
  1472.       )
  1473.     )
  1474.     (23): Detect(
  1475.       (cv2): ModuleList(
  1476.         (0): Sequential(
  1477.           (0): Conv(
  1478.             (quant): QuantStub(
  1479.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1480.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1481.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1482.               )
  1483.             )
  1484.             (dequant): DeQuantStub()
  1485.             (conv): ConvReLU2d(
  1486.               64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1487.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1488.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1489.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1490.               )
  1491.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1492.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1493.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1494.               )
  1495.             )
  1496.             (bn): Identity()
  1497.             (act): Identity()
  1498.           )
  1499.           (1): Conv(
  1500.             (quant): QuantStub(
  1501.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1502.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1503.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1504.               )
  1505.             )
  1506.             (dequant): DeQuantStub()
  1507.             (conv): ConvReLU2d(
  1508.               64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1509.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1510.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1511.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1512.               )
  1513.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1514.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1515.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1516.               )
  1517.             )
  1518.             (bn): Identity()
  1519.             (act): Identity()
  1520.           )
  1521.           (2): Conv2d(
  1522.             64, 64, kernel_size=(1, 1), stride=(1, 1)
  1523.             (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1524.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1525.               (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1526.             )
  1527.             (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1528.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1529.               (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1530.             )
  1531.           )
  1532.         )
  1533.         (1): Sequential(
  1534.           (0): Conv(
  1535.             (quant): QuantStub(
  1536.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1537.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1538.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1539.               )
  1540.             )
  1541.             (dequant): DeQuantStub()
  1542.             (conv): ConvReLU2d(
  1543.               128, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1544.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1545.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1546.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1547.               )
  1548.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1549.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1550.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1551.               )
  1552.             )
  1553.             (bn): Identity()
  1554.             (act): Identity()
  1555.           )
  1556.           (1): Conv(
  1557.             (quant): QuantStub(
  1558.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1559.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1560.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1561.               )
  1562.             )
  1563.             (dequant): DeQuantStub()
  1564.             (conv): ConvReLU2d(
  1565.               64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1566.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1567.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1568.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1569.               )
  1570.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1571.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1572.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1573.               )
  1574.             )
  1575.             (bn): Identity()
  1576.             (act): Identity()
  1577.           )
  1578.           (2): Conv2d(
  1579.             64, 64, kernel_size=(1, 1), stride=(1, 1)
  1580.             (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1581.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1582.               (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1583.             )
  1584.             (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1585.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1586.               (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1587.             )
  1588.           )
  1589.         )
  1590.         (2): Sequential(
  1591.           (0): Conv(
  1592.             (quant): QuantStub(
  1593.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1594.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1595.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1596.               )
  1597.             )
  1598.             (dequant): DeQuantStub()
  1599.             (conv): ConvReLU2d(
  1600.               256, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1601.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1602.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1603.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1604.               )
  1605.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1606.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1607.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1608.               )
  1609.             )
  1610.             (bn): Identity()
  1611.             (act): Identity()
  1612.           )
  1613.           (1): Conv(
  1614.             (quant): QuantStub(
  1615.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1616.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1617.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1618.               )
  1619.             )
  1620.             (dequant): DeQuantStub()
  1621.             (conv): ConvReLU2d(
  1622.               64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
  1623.               (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1624.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1625.                 (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1626.               )
  1627.               (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1628.                 fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1629.                 (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1630.               )
  1631.             )
  1632.             (bn): Identity()
  1633.             (act): Identity()
  1634.           )
  1635.           (2): Conv2d(
  1636.             64, 64, kernel_size=(1, 1), stride=(1, 1)
  1637.             (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1638.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1639.               (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1640.             )
  1641.             (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1642.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1643.               (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1644.             )
  1645.           )
  1646.         )
  1647.       )
  1648.       (cv3): ModuleList(
  1649.         (0): Sequential(
  1650.           (0): Sequential(
  1651.             (0): DWConv(
  1652.               (quant): QuantStub(
  1653.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1654.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1655.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1656.                 )
  1657.               )
  1658.               (dequant): DeQuantStub()
  1659.               (conv): ConvReLU2d(
  1660.                 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64
  1661.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1662.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1663.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1664.                 )
  1665.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1666.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1667.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1668.                 )
  1669.               )
  1670.               (bn): Identity()
  1671.               (act): Identity()
  1672.             )
  1673.             (1): Conv(
  1674.               (quant): QuantStub(
  1675.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1676.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1677.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1678.                 )
  1679.               )
  1680.               (dequant): DeQuantStub()
  1681.               (conv): ConvReLU2d(
  1682.                 64, 64, kernel_size=(1, 1), stride=(1, 1)
  1683.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1684.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1685.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1686.                 )
  1687.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1688.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1689.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1690.                 )
  1691.               )
  1692.               (bn): Identity()
  1693.               (act): Identity()
  1694.             )
  1695.           )
  1696.           (1): Sequential(
  1697.             (0): DWConv(
  1698.               (quant): QuantStub(
  1699.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1700.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1701.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1702.                 )
  1703.               )
  1704.               (dequant): DeQuantStub()
  1705.               (conv): ConvReLU2d(
  1706.                 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64
  1707.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1708.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1709.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1710.                 )
  1711.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1712.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1713.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1714.                 )
  1715.               )
  1716.               (bn): Identity()
  1717.               (act): Identity()
  1718.             )
  1719.             (1): Conv(
  1720.               (quant): QuantStub(
  1721.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1722.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1723.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1724.                 )
  1725.               )
  1726.               (dequant): DeQuantStub()
  1727.               (conv): ConvReLU2d(
  1728.                 64, 64, kernel_size=(1, 1), stride=(1, 1)
  1729.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1730.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1731.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1732.                 )
  1733.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1734.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1735.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1736.                 )
  1737.               )
  1738.               (bn): Identity()
  1739.               (act): Identity()
  1740.             )
  1741.           )
  1742.           (2): Conv2d(
  1743.             64, 2, kernel_size=(1, 1), stride=(1, 1)
  1744.             (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1745.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1746.               (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1747.             )
  1748.             (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1749.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1750.               (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1751.             )
  1752.           )
  1753.         )
  1754.         (1): Sequential(
  1755.           (0): Sequential(
  1756.             (0): DWConv(
  1757.               (quant): QuantStub(
  1758.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1759.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1760.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1761.                 )
  1762.               )
  1763.               (dequant): DeQuantStub()
  1764.               (conv): ConvReLU2d(
  1765.                 128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128
  1766.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1767.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1768.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1769.                 )
  1770.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1771.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1772.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1773.                 )
  1774.               )
  1775.               (bn): Identity()
  1776.               (act): Identity()
  1777.             )
  1778.             (1): Conv(
  1779.               (quant): QuantStub(
  1780.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1781.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1782.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1783.                 )
  1784.               )
  1785.               (dequant): DeQuantStub()
  1786.               (conv): ConvReLU2d(
  1787.                 128, 64, kernel_size=(1, 1), stride=(1, 1)
  1788.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1789.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1790.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1791.                 )
  1792.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1793.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1794.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1795.                 )
  1796.               )
  1797.               (bn): Identity()
  1798.               (act): Identity()
  1799.             )
  1800.           )
  1801.           (1): Sequential(
  1802.             (0): DWConv(
  1803.               (quant): QuantStub(
  1804.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1805.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1806.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1807.                 )
  1808.               )
  1809.               (dequant): DeQuantStub()
  1810.               (conv): ConvReLU2d(
  1811.                 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64
  1812.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1813.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1814.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1815.                 )
  1816.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1817.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1818.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1819.                 )
  1820.               )
  1821.               (bn): Identity()
  1822.               (act): Identity()
  1823.             )
  1824.             (1): Conv(
  1825.               (quant): QuantStub(
  1826.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1827.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1828.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1829.                 )
  1830.               )
  1831.               (dequant): DeQuantStub()
  1832.               (conv): ConvReLU2d(
  1833.                 64, 64, kernel_size=(1, 1), stride=(1, 1)
  1834.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1835.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1836.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1837.                 )
  1838.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1839.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1840.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1841.                 )
  1842.               )
  1843.               (bn): Identity()
  1844.               (act): Identity()
  1845.             )
  1846.           )
  1847.           (2): Conv2d(
  1848.             64, 2, kernel_size=(1, 1), stride=(1, 1)
  1849.             (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1850.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1851.               (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1852.             )
  1853.             (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1854.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1855.               (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1856.             )
  1857.           )
  1858.         )
  1859.         (2): Sequential(
  1860.           (0): Sequential(
  1861.             (0): DWConv(
  1862.               (quant): QuantStub(
  1863.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1864.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1865.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1866.                 )
  1867.               )
  1868.               (dequant): DeQuantStub()
  1869.               (conv): ConvReLU2d(
  1870.                 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256
  1871.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1872.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1873.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1874.                 )
  1875.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1876.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1877.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1878.                 )
  1879.               )
  1880.               (bn): Identity()
  1881.               (act): Identity()
  1882.             )
  1883.             (1): Conv(
  1884.               (quant): QuantStub(
  1885.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1886.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1887.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1888.                 )
  1889.               )
  1890.               (dequant): DeQuantStub()
  1891.               (conv): ConvReLU2d(
  1892.                 256, 64, kernel_size=(1, 1), stride=(1, 1)
  1893.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1894.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1895.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1896.                 )
  1897.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1898.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1899.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1900.                 )
  1901.               )
  1902.               (bn): Identity()
  1903.               (act): Identity()
  1904.             )
  1905.           )
  1906.           (1): Sequential(
  1907.             (0): DWConv(
  1908.               (quant): QuantStub(
  1909.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1910.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1911.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1912.                 )
  1913.               )
  1914.               (dequant): DeQuantStub()
  1915.               (conv): ConvReLU2d(
  1916.                 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64
  1917.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1918.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1919.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1920.                 )
  1921.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1922.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1923.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1924.                 )
  1925.               )
  1926.               (bn): Identity()
  1927.               (act): Identity()
  1928.             )
  1929.             (1): Conv(
  1930.               (quant): QuantStub(
  1931.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1932.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1933.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1934.                 )
  1935.               )
  1936.               (dequant): DeQuantStub()
  1937.               (conv): ConvReLU2d(
  1938.                 64, 64, kernel_size=(1, 1), stride=(1, 1)
  1939.                 (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1940.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1941.                   (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1942.                 )
  1943.                 (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1944.                   fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1945.                   (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1946.                 )
  1947.               )
  1948.               (bn): Identity()
  1949.               (act): Identity()
  1950.             )
  1951.           )
  1952.           (2): Conv2d(
  1953.             64, 2, kernel_size=(1, 1), stride=(1, 1)
  1954.             (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1955.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1956.               (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1957.             )
  1958.             (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1959.               fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1960.               (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1961.             )
  1962.           )
  1963.         )
  1964.       )
  1965.       (dfl): DFL(
  1966.         (conv): Conv2d(
  1967.           16, 1, kernel_size=(1, 1), stride=(1, 1), bias=False
  1968.           (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
  1969.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
  1970.             (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
  1971.           )
  1972.           (activation_post_process): FusedMovingAvgObsFakeQuantize(
  1973.             fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
  1974.             (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
  1975.           )
  1976.         )
  1977.       )
  1978.     )
  1979.   )
  1980. )
Advertisement
Add Comment
Please, Sign In to add comment