resnet-50 nnvm ir

gistfile1.txt

Graph(%input_tensor,
      %resnet_model/conv2d/kernel,
      %resnet_model/batch_normalization/gamma,
      %resnet_model/batch_normalization/beta,
      %resnet_model/batch_normalization/moving_mean,
      %resnet_model/batch_normalization/moving_variance,
      %resnet_model/conv2d_2/kernel,
      %resnet_model/batch_normalization_2/gamma,
      %resnet_model/batch_normalization_2/beta,
      %resnet_model/batch_normalization_2/moving_mean,
      %resnet_model/batch_normalization_2/moving_variance,
      %resnet_model/conv2d_3/kernel,
      %resnet_model/batch_normalization_3/gamma,
      %resnet_model/batch_normalization_3/beta,
      %resnet_model/batch_normalization_3/moving_mean,
      %resnet_model/batch_normalization_3/moving_variance,
      %resnet_model/conv2d_4/kernel,
      %resnet_model/batch_normalization_4/gamma,
      %resnet_model/batch_normalization_4/beta,
      %resnet_model/batch_normalization_4/moving_mean,
      %resnet_model/batch_normalization_4/moving_variance,
      %resnet_model/conv2d_1/kernel,
      %resnet_model/batch_normalization_1/gamma,
      %resnet_model/batch_normalization_1/beta,
      %resnet_model/batch_normalization_1/moving_mean,
      %resnet_model/batch_normalization_1/moving_variance,
      %resnet_model/conv2d_5/kernel,
      %resnet_model/batch_normalization_5/gamma,
      %resnet_model/batch_normalization_5/beta,
      %resnet_model/batch_normalization_5/moving_mean,
      %resnet_model/batch_normalization_5/moving_variance,
      %resnet_model/conv2d_6/kernel,
      %resnet_model/batch_normalization_6/gamma,
      %resnet_model/batch_normalization_6/beta,
      %resnet_model/batch_normalization_6/moving_mean,
      %resnet_model/batch_normalization_6/moving_variance,
      %resnet_model/conv2d_7/kernel,
      %resnet_model/batch_normalization_7/gamma,
      %resnet_model/batch_normalization_7/beta,
      %resnet_model/batch_normalization_7/moving_mean,
      %resnet_model/batch_normalization_7/moving_variance,
      %resnet_model/conv2d_8/kernel,
      %resnet_model/batch_normalization_8/gamma,
      %resnet_model/batch_normalization_8/beta,
      %resnet_model/batch_normalization_8/moving_mean,
      %resnet_model/batch_normalization_8/moving_variance,
      %resnet_model/conv2d_9/kernel,
      %resnet_model/batch_normalization_9/gamma,
      %resnet_model/batch_normalization_9/beta,
      %resnet_model/batch_normalization_9/moving_mean,
      %resnet_model/batch_normalization_9/moving_variance,
      %resnet_model/conv2d_10/kernel,
      %resnet_model/batch_normalization_10/gamma,
      %resnet_model/batch_normalization_10/beta,
      %resnet_model/batch_normalization_10/moving_mean,
      %resnet_model/batch_normalization_10/moving_variance,
      %resnet_model/conv2d_12/kernel,
      %resnet_model/batch_normalization_12/gamma,
      %resnet_model/batch_normalization_12/beta,
      %resnet_model/batch_normalization_12/moving_mean,
      %resnet_model/batch_normalization_12/moving_variance,
      %resnet_model/conv2d_13/kernel,
      %resnet_model/batch_normalization_13/gamma,
      %resnet_model/batch_normalization_13/beta,
      %resnet_model/batch_normalization_13/moving_mean,
      %resnet_model/batch_normalization_13/moving_variance,
      %resnet_model/conv2d_14/kernel,
      %resnet_model/batch_normalization_14/gamma,
      %resnet_model/batch_normalization_14/beta,
      %resnet_model/batch_normalization_14/moving_mean,
      %resnet_model/batch_normalization_14/moving_variance,
      %resnet_model/conv2d_11/kernel,
      %resnet_model/batch_normalization_11/gamma,
      %resnet_model/batch_normalization_11/beta,
      %resnet_model/batch_normalization_11/moving_mean,
      %resnet_model/batch_normalization_11/moving_variance,
      %resnet_model/conv2d_15/kernel,
      %resnet_model/batch_normalization_15/gamma,
      %resnet_model/batch_normalization_15/beta,
      %resnet_model/batch_normalization_15/moving_mean,
      %resnet_model/batch_normalization_15/moving_variance,
      %resnet_model/conv2d_16/kernel,
      %resnet_model/batch_normalization_16/gamma,
      %resnet_model/batch_normalization_16/beta,
      %resnet_model/batch_normalization_16/moving_mean,
      %resnet_model/batch_normalization_16/moving_variance,
      %resnet_model/conv2d_17/kernel,
      %resnet_model/batch_normalization_17/gamma,
      %resnet_model/batch_normalization_17/beta,
      %resnet_model/batch_normalization_17/moving_mean,
      %resnet_model/batch_normalization_17/moving_variance,
      %resnet_model/conv2d_18/kernel,
      %resnet_model/batch_normalization_18/gamma,
      %resnet_model/batch_normalization_18/beta,
      %resnet_model/batch_normalization_18/moving_mean,
      %resnet_model/batch_normalization_18/moving_variance,
      %resnet_model/conv2d_19/kernel,
      %resnet_model/batch_normalization_19/gamma,
      %resnet_model/batch_normalization_19/beta,
      %resnet_model/batch_normalization_19/moving_mean,
      %resnet_model/batch_normalization_19/moving_variance,
      %resnet_model/conv2d_20/kernel,
      %resnet_model/batch_normalization_20/gamma,
      %resnet_model/batch_normalization_20/beta,
      %resnet_model/batch_normalization_20/moving_mean,
      %resnet_model/batch_normalization_20/moving_variance,
      %resnet_model/conv2d_21/kernel,
      %resnet_model/batch_normalization_21/gamma,
      %resnet_model/batch_normalization_21/beta,
      %resnet_model/batch_normalization_21/moving_mean,
      %resnet_model/batch_normalization_21/moving_variance,
      %resnet_model/conv2d_22/kernel,
      %resnet_model/batch_normalization_22/gamma,
      %resnet_model/batch_normalization_22/beta,
      %resnet_model/batch_normalization_22/moving_mean,
      %resnet_model/batch_normalization_22/moving_variance,
      %resnet_model/conv2d_23/kernel,
      %resnet_model/batch_normalization_23/gamma,
      %resnet_model/batch_normalization_23/beta,
      %resnet_model/batch_normalization_23/moving_mean,
      %resnet_model/batch_normalization_23/moving_variance,
      %resnet_model/conv2d_25/kernel,
      %resnet_model/batch_normalization_25/gamma,
      %resnet_model/batch_normalization_25/beta,
      %resnet_model/batch_normalization_25/moving_mean,
      %resnet_model/batch_normalization_25/moving_variance,
      %resnet_model/conv2d_26/kernel,
      %resnet_model/batch_normalization_26/gamma,
      %resnet_model/batch_normalization_26/beta,
      %resnet_model/batch_normalization_26/moving_mean,
      %resnet_model/batch_normalization_26/moving_variance,
      %resnet_model/conv2d_27/kernel,
      %resnet_model/batch_normalization_27/gamma,
      %resnet_model/batch_normalization_27/beta,
      %resnet_model/batch_normalization_27/moving_mean,
      %resnet_model/batch_normalization_27/moving_variance,
      %resnet_model/conv2d_24/kernel,
      %resnet_model/batch_normalization_24/gamma,
      %resnet_model/batch_normalization_24/beta,
      %resnet_model/batch_normalization_24/moving_mean,
      %resnet_model/batch_normalization_24/moving_variance,
      %resnet_model/conv2d_28/kernel,
      %resnet_model/batch_normalization_28/gamma,
      %resnet_model/batch_normalization_28/beta,
      %resnet_model/batch_normalization_28/moving_mean,
      %resnet_model/batch_normalization_28/moving_variance,
      %resnet_model/conv2d_29/kernel,
      %resnet_model/batch_normalization_29/gamma,
      %resnet_model/batch_normalization_29/beta,
      %resnet_model/batch_normalization_29/moving_mean,
      %resnet_model/batch_normalization_29/moving_variance,
      %resnet_model/conv2d_30/kernel,
      %resnet_model/batch_normalization_30/gamma,
      %resnet_model/batch_normalization_30/beta,
      %resnet_model/batch_normalization_30/moving_mean,
      %resnet_model/batch_normalization_30/moving_variance,
      %resnet_model/conv2d_31/kernel,
      %resnet_model/batch_normalization_31/gamma,
      %resnet_model/batch_normalization_31/beta,
      %resnet_model/batch_normalization_31/moving_mean,
      %resnet_model/batch_normalization_31/moving_variance,
      %resnet_model/conv2d_32/kernel,
      %resnet_model/batch_normalization_32/gamma,
      %resnet_model/batch_normalization_32/beta,
      %resnet_model/batch_normalization_32/moving_mean,
      %resnet_model/batch_normalization_32/moving_variance,
      %resnet_model/conv2d_33/kernel,
      %resnet_model/batch_normalization_33/gamma,
      %resnet_model/batch_normalization_33/beta,
      %resnet_model/batch_normalization_33/moving_mean,
      %resnet_model/batch_normalization_33/moving_variance,
      %resnet_model/conv2d_34/kernel,
      %resnet_model/batch_normalization_34/gamma,
      %resnet_model/batch_normalization_34/beta,
      %resnet_model/batch_normalization_34/moving_mean,
      %resnet_model/batch_normalization_34/moving_variance,
      %resnet_model/conv2d_35/kernel,
      %resnet_model/batch_normalization_35/gamma,
      %resnet_model/batch_normalization_35/beta,
      %resnet_model/batch_normalization_35/moving_mean,
      %resnet_model/batch_normalization_35/moving_variance,
      %resnet_model/conv2d_36/kernel,
      %resnet_model/batch_normalization_36/gamma,
      %resnet_model/batch_normalization_36/beta,
      %resnet_model/batch_normalization_36/moving_mean,
      %resnet_model/batch_normalization_36/moving_variance,
      %resnet_model/conv2d_37/kernel,
      %resnet_model/batch_normalization_37/gamma,
      %resnet_model/batch_normalization_37/beta,
      %resnet_model/batch_normalization_37/moving_mean,
      %resnet_model/batch_normalization_37/moving_variance,
      %resnet_model/conv2d_38/kernel,
      %resnet_model/batch_normalization_38/gamma,
      %resnet_model/batch_normalization_38/beta,
      %resnet_model/batch_normalization_38/moving_mean,
      %resnet_model/batch_normalization_38/moving_variance,
      %resnet_model/conv2d_39/kernel,
      %resnet_model/batch_normalization_39/gamma,
      %resnet_model/batch_normalization_39/beta,
      %resnet_model/batch_normalization_39/moving_mean,
      %resnet_model/batch_normalization_39/moving_variance,
      %resnet_model/conv2d_40/kernel,
      %resnet_model/batch_normalization_40/gamma,
      %resnet_model/batch_normalization_40/beta,
      %resnet_model/batch_normalization_40/moving_mean,
      %resnet_model/batch_normalization_40/moving_variance,
      %resnet_model/conv2d_41/kernel,
      %resnet_model/batch_normalization_41/gamma,
      %resnet_model/batch_normalization_41/beta,
      %resnet_model/batch_normalization_41/moving_mean,
      %resnet_model/batch_normalization_41/moving_variance,
      %resnet_model/conv2d_42/kernel,
      %resnet_model/batch_normalization_42/gamma,
      %resnet_model/batch_normalization_42/beta,
      %resnet_model/batch_normalization_42/moving_mean,
      %resnet_model/batch_normalization_42/moving_variance,
      %resnet_model/conv2d_44/kernel,
      %resnet_model/batch_normalization_44/gamma,
      %resnet_model/batch_normalization_44/beta,
      %resnet_model/batch_normalization_44/moving_mean,
      %resnet_model/batch_normalization_44/moving_variance,
      %resnet_model/conv2d_45/kernel,
      %resnet_model/batch_normalization_45/gamma,
      %resnet_model/batch_normalization_45/beta,
      %resnet_model/batch_normalization_45/moving_mean,
      %resnet_model/batch_normalization_45/moving_variance,
      %resnet_model/conv2d_46/kernel,
      %resnet_model/batch_normalization_46/gamma,
      %resnet_model/batch_normalization_46/beta,
      %resnet_model/batch_normalization_46/moving_mean,
      %resnet_model/batch_normalization_46/moving_variance,
      %resnet_model/conv2d_43/kernel,
      %resnet_model/batch_normalization_43/gamma,
      %resnet_model/batch_normalization_43/beta,
      %resnet_model/batch_normalization_43/moving_mean,
      %resnet_model/batch_normalization_43/moving_variance,
      %resnet_model/conv2d_47/kernel,
      %resnet_model/batch_normalization_47/gamma,
      %resnet_model/batch_normalization_47/beta,
      %resnet_model/batch_normalization_47/moving_mean,
      %resnet_model/batch_normalization_47/moving_variance,
      %resnet_model/conv2d_48/kernel,
      %resnet_model/batch_normalization_48/gamma,
      %resnet_model/batch_normalization_48/beta,
      %resnet_model/batch_normalization_48/moving_mean,
      %resnet_model/batch_normalization_48/moving_variance,
      %resnet_model/conv2d_49/kernel,
      %resnet_model/batch_normalization_49/gamma,
      %resnet_model/batch_normalization_49/beta,
      %resnet_model/batch_normalization_49/moving_mean,
      %resnet_model/batch_normalization_49/moving_variance,
      %resnet_model/conv2d_50/kernel,
      %resnet_model/batch_normalization_50/gamma,
      %resnet_model/batch_normalization_50/beta,
      %resnet_model/batch_normalization_50/moving_mean,
      %resnet_model/batch_normalization_50/moving_variance,
      %resnet_model/conv2d_51/kernel,
      %resnet_model/batch_normalization_51/gamma,
      %resnet_model/batch_normalization_51/beta,
      %resnet_model/batch_normalization_51/moving_mean,
      %resnet_model/batch_normalization_51/moving_variance,
      %resnet_model/conv2d_52/kernel,
      %resnet_model/batch_normalization_52/gamma,
      %resnet_model/batch_normalization_52/beta,
      %resnet_model/batch_normalization_52/moving_mean,
      %resnet_model/batch_normalization_52/moving_variance,
      %resnet_model/dense/kernel,
      %resnet_model/dense/bias) {
  %1 = transpose(%input_tensor, axes='(0, 3, 1, 2)')
  %2 = pad(%1, pad_value='0', pad_width='((0, 0), (0, 0), (3, 3), (3, 3))')
  %4 = transpose(%resnet_model/conv2d/kernel, axes='(3, 2, 0, 1)')
  %5 = conv2d(%2, %4, kernel_layout='OIHW', use_bias='False', channels='64', kernel_size='(7, 7)', strides='(2, 2)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %10 = batch_norm(%5, %resnet_model/batch_normalization/gamma, %resnet_model/batch_normalization/beta, %resnet_model/batch_normalization/moving_mean, %resnet_model/batch_normalization/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %11 = relu(%10.0)
  %12 = max_pool2d(%11, pool_size='(3, 3)', strides='(2, 2)', ceil_mode='False', padding='[0, 0, 1, 1]', layout='NCHW')
  %13 = pad(%12, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %15 = transpose(%resnet_model/conv2d_2/kernel, axes='(3, 2, 0, 1)')
  %16 = conv2d(%13, %15, kernel_layout='OIHW', use_bias='False', channels='64', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %21 = batch_norm(%16, %resnet_model/batch_normalization_2/gamma, %resnet_model/batch_normalization_2/beta, %resnet_model/batch_normalization_2/moving_mean, %resnet_model/batch_normalization_2/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %22 = relu(%21.0)
  %23 = pad(%22, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %25 = transpose(%resnet_model/conv2d_3/kernel, axes='(3, 2, 0, 1)')
  %26 = conv2d(%23, %25, kernel_layout='OIHW', use_bias='False', channels='64', kernel_size='(3, 3)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %31 = batch_norm(%26, %resnet_model/batch_normalization_3/gamma, %resnet_model/batch_normalization_3/beta, %resnet_model/batch_normalization_3/moving_mean, %resnet_model/batch_normalization_3/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %32 = relu(%31.0)
  %33 = pad(%32, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %35 = transpose(%resnet_model/conv2d_4/kernel, axes='(3, 2, 0, 1)')
  %36 = conv2d(%33, %35, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %41 = batch_norm(%36, %resnet_model/batch_normalization_4/gamma, %resnet_model/batch_normalization_4/beta, %resnet_model/batch_normalization_4/moving_mean, %resnet_model/batch_normalization_4/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %42 = pad(%12, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %44 = transpose(%resnet_model/conv2d_1/kernel, axes='(3, 2, 0, 1)')
  %45 = conv2d(%42, %44, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %50 = batch_norm(%45, %resnet_model/batch_normalization_1/gamma, %resnet_model/batch_normalization_1/beta, %resnet_model/batch_normalization_1/moving_mean, %resnet_model/batch_normalization_1/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %51 = broadcast_add(%41.0, %50.0)
  %52 = relu(%51)
  %53 = pad(%52, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %55 = transpose(%resnet_model/conv2d_5/kernel, axes='(3, 2, 0, 1)')
  %56 = conv2d(%53, %55, kernel_layout='OIHW', use_bias='False', channels='64', kernel_size='(1, 1)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %61 = batch_norm(%56, %resnet_model/batch_normalization_5/gamma, %resnet_model/batch_normalization_5/beta, %resnet_model/batch_normalization_5/moving_mean, %resnet_model/batch_normalization_5/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %62 = relu(%61.0)
  %63 = pad(%62, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %65 = transpose(%resnet_model/conv2d_6/kernel, axes='(3, 2, 0, 1)')
  %66 = conv2d(%63, %65, kernel_layout='OIHW', use_bias='False', channels='64', kernel_size='(3, 3)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %71 = batch_norm(%66, %resnet_model/batch_normalization_6/gamma, %resnet_model/batch_normalization_6/beta, %resnet_model/batch_normalization_6/moving_mean, %resnet_model/batch_normalization_6/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %72 = relu(%71.0)
  %73 = pad(%72, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %75 = transpose(%resnet_model/conv2d_7/kernel, axes='(3, 2, 0, 1)')
  %76 = conv2d(%73, %75, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %81 = batch_norm(%76, %resnet_model/batch_normalization_7/gamma, %resnet_model/batch_normalization_7/beta, %resnet_model/batch_normalization_7/moving_mean, %resnet_model/batch_normalization_7/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %82 = broadcast_add(%81.0, %52)
  %83 = relu(%82)
  %84 = pad(%83, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %86 = transpose(%resnet_model/conv2d_8/kernel, axes='(3, 2, 0, 1)')
  %87 = conv2d(%84, %86, kernel_layout='OIHW', use_bias='False', channels='64', kernel_size='(1, 1)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %92 = batch_norm(%87, %resnet_model/batch_normalization_8/gamma, %resnet_model/batch_normalization_8/beta, %resnet_model/batch_normalization_8/moving_mean, %resnet_model/batch_normalization_8/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %93 = relu(%92.0)
  %94 = pad(%93, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %96 = transpose(%resnet_model/conv2d_9/kernel, axes='(3, 2, 0, 1)')
  %97 = conv2d(%94, %96, kernel_layout='OIHW', use_bias='False', channels='64', kernel_size='(3, 3)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %102 = batch_norm(%97, %resnet_model/batch_normalization_9/gamma, %resnet_model/batch_normalization_9/beta, %resnet_model/batch_normalization_9/moving_mean, %resnet_model/batch_normalization_9/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %103 = relu(%102.0)
  %104 = pad(%103, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %106 = transpose(%resnet_model/conv2d_10/kernel, axes='(3, 2, 0, 1)')
  %107 = conv2d(%104, %106, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %112 = batch_norm(%107, %resnet_model/batch_normalization_10/gamma, %resnet_model/batch_normalization_10/beta, %resnet_model/batch_normalization_10/moving_mean, %resnet_model/batch_normalization_10/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %113 = broadcast_add(%112.0, %83)
  %114 = relu(%113)
  %115 = pad(%114, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %117 = transpose(%resnet_model/conv2d_12/kernel, axes='(3, 2, 0, 1)')
  %118 = conv2d(%115, %117, kernel_layout='OIHW', use_bias='False', channels='128', kernel_size='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW', strides='(1, 1)')
  %123 = batch_norm(%118, %resnet_model/batch_normalization_12/gamma, %resnet_model/batch_normalization_12/beta, %resnet_model/batch_normalization_12/moving_mean, %resnet_model/batch_normalization_12/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %124 = relu(%123.0)
  %125 = pad(%124, pad_value='0', pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %127 = transpose(%resnet_model/conv2d_13/kernel, axes='(3, 2, 0, 1)')
  %128 = conv2d(%125, %127, kernel_layout='OIHW', use_bias='False', channels='128', kernel_size='(3, 3)', strides='(2, 2)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %133 = batch_norm(%128, %resnet_model/batch_normalization_13/gamma, %resnet_model/batch_normalization_13/beta, %resnet_model/batch_normalization_13/moving_mean, %resnet_model/batch_normalization_13/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %134 = relu(%133.0)
  %135 = pad(%134, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %137 = transpose(%resnet_model/conv2d_14/kernel, axes='(3, 2, 0, 1)')
  %138 = conv2d(%135, %137, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %143 = batch_norm(%138, %resnet_model/batch_normalization_14/gamma, %resnet_model/batch_normalization_14/beta, %resnet_model/batch_normalization_14/moving_mean, %resnet_model/batch_normalization_14/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %144 = pad(%114, pad_value='0', pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %146 = transpose(%resnet_model/conv2d_11/kernel, axes='(3, 2, 0, 1)')
  %147 = conv2d(%144, %146, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(1, 1)', strides='(2, 2)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %152 = batch_norm(%147, %resnet_model/batch_normalization_11/gamma, %resnet_model/batch_normalization_11/beta, %resnet_model/batch_normalization_11/moving_mean, %resnet_model/batch_normalization_11/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %153 = broadcast_add(%143.0, %152.0)
  %154 = relu(%153)
  %155 = pad(%154, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %157 = transpose(%resnet_model/conv2d_15/kernel, axes='(3, 2, 0, 1)')
  %158 = conv2d(%155, %157, kernel_layout='OIHW', use_bias='False', channels='128', kernel_size='(1, 1)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %163 = batch_norm(%158, %resnet_model/batch_normalization_15/gamma, %resnet_model/batch_normalization_15/beta, %resnet_model/batch_normalization_15/moving_mean, %resnet_model/batch_normalization_15/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %164 = relu(%163.0)
  %165 = pad(%164, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %167 = transpose(%resnet_model/conv2d_16/kernel, axes='(3, 2, 0, 1)')
  %168 = conv2d(%165, %167, kernel_layout='OIHW', use_bias='False', channels='128', kernel_size='(3, 3)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %173 = batch_norm(%168, %resnet_model/batch_normalization_16/gamma, %resnet_model/batch_normalization_16/beta, %resnet_model/batch_normalization_16/moving_mean, %resnet_model/batch_normalization_16/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %174 = relu(%173.0)
  %175 = pad(%174, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %177 = transpose(%resnet_model/conv2d_17/kernel, axes='(3, 2, 0, 1)')
  %178 = conv2d(%175, %177, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %183 = batch_norm(%178, %resnet_model/batch_normalization_17/gamma, %resnet_model/batch_normalization_17/beta, %resnet_model/batch_normalization_17/moving_mean, %resnet_model/batch_normalization_17/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %184 = broadcast_add(%183.0, %154)
  %185 = relu(%184)
  %186 = pad(%185, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %188 = transpose(%resnet_model/conv2d_18/kernel, axes='(3, 2, 0, 1)')
  %189 = conv2d(%186, %188, kernel_layout='OIHW', use_bias='False', channels='128', kernel_size='(1, 1)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %194 = batch_norm(%189, %resnet_model/batch_normalization_18/gamma, %resnet_model/batch_normalization_18/beta, %resnet_model/batch_normalization_18/moving_mean, %resnet_model/batch_normalization_18/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %195 = relu(%194.0)
  %196 = pad(%195, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %198 = transpose(%resnet_model/conv2d_19/kernel, axes='(3, 2, 0, 1)')
  %199 = conv2d(%196, %198, kernel_layout='OIHW', use_bias='False', channels='128', kernel_size='(3, 3)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %204 = batch_norm(%199, %resnet_model/batch_normalization_19/gamma, %resnet_model/batch_normalization_19/beta, %resnet_model/batch_normalization_19/moving_mean, %resnet_model/batch_normalization_19/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %205 = relu(%204.0)
  %206 = pad(%205, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %208 = transpose(%resnet_model/conv2d_20/kernel, axes='(3, 2, 0, 1)')
  %209 = conv2d(%206, %208, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(1, 1)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %214 = batch_norm(%209, %resnet_model/batch_normalization_20/gamma, %resnet_model/batch_normalization_20/beta, %resnet_model/batch_normalization_20/moving_mean, %resnet_model/batch_normalization_20/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %215 = broadcast_add(%214.0, %185)
  %216 = relu(%215)
  %217 = pad(%216, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %219 = transpose(%resnet_model/conv2d_21/kernel, axes='(3, 2, 0, 1)')
  %220 = conv2d(%217, %219, kernel_layout='OIHW', use_bias='False', channels='128', kernel_size='(1, 1)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %225 = batch_norm(%220, %resnet_model/batch_normalization_21/gamma, %resnet_model/batch_normalization_21/beta, %resnet_model/batch_normalization_21/moving_mean, %resnet_model/batch_normalization_21/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %226 = relu(%225.0)
  %227 = pad(%226, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %229 = transpose(%resnet_model/conv2d_22/kernel, axes='(3, 2, 0, 1)')
  %230 = conv2d(%227, %229, kernel_layout='OIHW', use_bias='False', channels='128', kernel_size='(3, 3)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW', strides='(1, 1)')
  %235 = batch_norm(%230, %resnet_model/batch_normalization_22/gamma, %resnet_model/batch_normalization_22/beta, %resnet_model/batch_normalization_22/moving_mean, %resnet_model/batch_normalization_22/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %236 = relu(%235.0)
  %237 = pad(%236, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %239 = transpose(%resnet_model/conv2d_23/kernel, axes='(3, 2, 0, 1)')
  %240 = conv2d(%237, %239, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(1, 1)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %245 = batch_norm(%240, %resnet_model/batch_normalization_23/gamma, %resnet_model/batch_normalization_23/beta, %resnet_model/batch_normalization_23/moving_mean, %resnet_model/batch_normalization_23/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %246 = broadcast_add(%245.0, %216)
  %247 = relu(%246)
  %248 = pad(%247, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %250 = transpose(%resnet_model/conv2d_25/kernel, axes='(3, 2, 0, 1)')
  %251 = conv2d(%248, %250, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %256 = batch_norm(%251, %resnet_model/batch_normalization_25/gamma, %resnet_model/batch_normalization_25/beta, %resnet_model/batch_normalization_25/moving_mean, %resnet_model/batch_normalization_25/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %257 = relu(%256.0)
  %258 = pad(%257, pad_value='0', pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %260 = transpose(%resnet_model/conv2d_26/kernel, axes='(3, 2, 0, 1)')
  %261 = conv2d(%258, %260, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(3, 3)', strides='(2, 2)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %266 = batch_norm(%261, %resnet_model/batch_normalization_26/gamma, %resnet_model/batch_normalization_26/beta, %resnet_model/batch_normalization_26/moving_mean, %resnet_model/batch_normalization_26/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %267 = relu(%266.0)
  %268 = pad(%267, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %270 = transpose(%resnet_model/conv2d_27/kernel, axes='(3, 2, 0, 1)')
  %271 = conv2d(%268, %270, kernel_layout='OIHW', use_bias='False', channels='1024', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %276 = batch_norm(%271, %resnet_model/batch_normalization_27/gamma, %resnet_model/batch_normalization_27/beta, %resnet_model/batch_normalization_27/moving_mean, %resnet_model/batch_normalization_27/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %277 = pad(%247, pad_value='0', pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %279 = transpose(%resnet_model/conv2d_24/kernel, axes='(3, 2, 0, 1)')
  %280 = conv2d(%277, %279, kernel_layout='OIHW', use_bias='False', channels='1024', kernel_size='(1, 1)', strides='(2, 2)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %285 = batch_norm(%280, %resnet_model/batch_normalization_24/gamma, %resnet_model/batch_normalization_24/beta, %resnet_model/batch_normalization_24/moving_mean, %resnet_model/batch_normalization_24/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %286 = broadcast_add(%276.0, %285.0)
  %287 = relu(%286)
  %288 = pad(%287, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %290 = transpose(%resnet_model/conv2d_28/kernel, axes='(3, 2, 0, 1)')
  %291 = conv2d(%288, %290, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %296 = batch_norm(%291, %resnet_model/batch_normalization_28/gamma, %resnet_model/batch_normalization_28/beta, %resnet_model/batch_normalization_28/moving_mean, %resnet_model/batch_normalization_28/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %297 = relu(%296.0)
  %298 = pad(%297, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %300 = transpose(%resnet_model/conv2d_29/kernel, axes='(3, 2, 0, 1)')
  %301 = conv2d(%298, %300, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(3, 3)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %306 = batch_norm(%301, %resnet_model/batch_normalization_29/gamma, %resnet_model/batch_normalization_29/beta, %resnet_model/batch_normalization_29/moving_mean, %resnet_model/batch_normalization_29/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %307 = relu(%306.0)
  %308 = pad(%307, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %310 = transpose(%resnet_model/conv2d_30/kernel, axes='(3, 2, 0, 1)')
  %311 = conv2d(%308, %310, kernel_layout='OIHW', use_bias='False', channels='1024', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %316 = batch_norm(%311, %resnet_model/batch_normalization_30/gamma, %resnet_model/batch_normalization_30/beta, %resnet_model/batch_normalization_30/moving_mean, %resnet_model/batch_normalization_30/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %317 = broadcast_add(%316.0, %287)
  %318 = relu(%317)
  %319 = pad(%318, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %321 = transpose(%resnet_model/conv2d_31/kernel, axes='(3, 2, 0, 1)')
  %322 = conv2d(%319, %321, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %327 = batch_norm(%322, %resnet_model/batch_normalization_31/gamma, %resnet_model/batch_normalization_31/beta, %resnet_model/batch_normalization_31/moving_mean, %resnet_model/batch_normalization_31/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %328 = relu(%327.0)
  %329 = pad(%328, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %331 = transpose(%resnet_model/conv2d_32/kernel, axes='(3, 2, 0, 1)')
  %332 = conv2d(%329, %331, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(3, 3)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %337 = batch_norm(%332, %resnet_model/batch_normalization_32/gamma, %resnet_model/batch_normalization_32/beta, %resnet_model/batch_normalization_32/moving_mean, %resnet_model/batch_normalization_32/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %338 = relu(%337.0)
  %339 = pad(%338, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %341 = transpose(%resnet_model/conv2d_33/kernel, axes='(3, 2, 0, 1)')
  %342 = conv2d(%339, %341, kernel_layout='OIHW', use_bias='False', channels='1024', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %347 = batch_norm(%342, %resnet_model/batch_normalization_33/gamma, %resnet_model/batch_normalization_33/beta, %resnet_model/batch_normalization_33/moving_mean, %resnet_model/batch_normalization_33/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %348 = broadcast_add(%347.0, %318)
  %349 = relu(%348)
  %350 = pad(%349, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %352 = transpose(%resnet_model/conv2d_34/kernel, axes='(3, 2, 0, 1)')
  %353 = conv2d(%350, %352, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %358 = batch_norm(%353, %resnet_model/batch_normalization_34/gamma, %resnet_model/batch_normalization_34/beta, %resnet_model/batch_normalization_34/moving_mean, %resnet_model/batch_normalization_34/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %359 = relu(%358.0)
  %360 = pad(%359, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %362 = transpose(%resnet_model/conv2d_35/kernel, axes='(3, 2, 0, 1)')
  %363 = conv2d(%360, %362, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(3, 3)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %368 = batch_norm(%363, %resnet_model/batch_normalization_35/gamma, %resnet_model/batch_normalization_35/beta, %resnet_model/batch_normalization_35/moving_mean, %resnet_model/batch_normalization_35/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %369 = relu(%368.0)
  %370 = pad(%369, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %372 = transpose(%resnet_model/conv2d_36/kernel, axes='(3, 2, 0, 1)')
  %373 = conv2d(%370, %372, kernel_layout='OIHW', use_bias='False', channels='1024', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %378 = batch_norm(%373, %resnet_model/batch_normalization_36/gamma, %resnet_model/batch_normalization_36/beta, %resnet_model/batch_normalization_36/moving_mean, %resnet_model/batch_normalization_36/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %379 = broadcast_add(%378.0, %349)
  %380 = relu(%379)
  %381 = pad(%380, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %383 = transpose(%resnet_model/conv2d_37/kernel, axes='(3, 2, 0, 1)')
  %384 = conv2d(%381, %383, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %389 = batch_norm(%384, %resnet_model/batch_normalization_37/gamma, %resnet_model/batch_normalization_37/beta, %resnet_model/batch_normalization_37/moving_mean, %resnet_model/batch_normalization_37/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %390 = relu(%389.0)
  %391 = pad(%390, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %393 = transpose(%resnet_model/conv2d_38/kernel, axes='(3, 2, 0, 1)')
  %394 = conv2d(%391, %393, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(3, 3)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %399 = batch_norm(%394, %resnet_model/batch_normalization_38/gamma, %resnet_model/batch_normalization_38/beta, %resnet_model/batch_normalization_38/moving_mean, %resnet_model/batch_normalization_38/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %400 = relu(%399.0)
  %401 = pad(%400, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %403 = transpose(%resnet_model/conv2d_39/kernel, axes='(3, 2, 0, 1)')
  %404 = conv2d(%401, %403, kernel_layout='OIHW', use_bias='False', channels='1024', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %409 = batch_norm(%404, %resnet_model/batch_normalization_39/gamma, %resnet_model/batch_normalization_39/beta, %resnet_model/batch_normalization_39/moving_mean, %resnet_model/batch_normalization_39/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %410 = broadcast_add(%409.0, %380)
  %411 = relu(%410)
  %412 = pad(%411, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %414 = transpose(%resnet_model/conv2d_40/kernel, axes='(3, 2, 0, 1)')
  %415 = conv2d(%412, %414, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %420 = batch_norm(%415, %resnet_model/batch_normalization_40/gamma, %resnet_model/batch_normalization_40/beta, %resnet_model/batch_normalization_40/moving_mean, %resnet_model/batch_normalization_40/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %421 = relu(%420.0)
  %422 = pad(%421, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %424 = transpose(%resnet_model/conv2d_41/kernel, axes='(3, 2, 0, 1)')
  %425 = conv2d(%422, %424, kernel_layout='OIHW', use_bias='False', channels='256', kernel_size='(3, 3)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %430 = batch_norm(%425, %resnet_model/batch_normalization_41/gamma, %resnet_model/batch_normalization_41/beta, %resnet_model/batch_normalization_41/moving_mean, %resnet_model/batch_normalization_41/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %431 = relu(%430.0)
  %432 = pad(%431, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %434 = transpose(%resnet_model/conv2d_42/kernel, axes='(3, 2, 0, 1)')
  %435 = conv2d(%432, %434, kernel_layout='OIHW', use_bias='False', channels='1024', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %440 = batch_norm(%435, %resnet_model/batch_normalization_42/gamma, %resnet_model/batch_normalization_42/beta, %resnet_model/batch_normalization_42/moving_mean, %resnet_model/batch_normalization_42/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %441 = broadcast_add(%440.0, %411)
  %442 = relu(%441)
  %443 = pad(%442, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %445 = transpose(%resnet_model/conv2d_44/kernel, axes='(3, 2, 0, 1)')
  %446 = conv2d(%443, %445, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %451 = batch_norm(%446, %resnet_model/batch_normalization_44/gamma, %resnet_model/batch_normalization_44/beta, %resnet_model/batch_normalization_44/moving_mean, %resnet_model/batch_normalization_44/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %452 = relu(%451.0)
  %453 = pad(%452, pad_value='0', pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %455 = transpose(%resnet_model/conv2d_45/kernel, axes='(3, 2, 0, 1)')
  %456 = conv2d(%453, %455, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(3, 3)', strides='(2, 2)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %461 = batch_norm(%456, %resnet_model/batch_normalization_45/gamma, %resnet_model/batch_normalization_45/beta, %resnet_model/batch_normalization_45/moving_mean, %resnet_model/batch_normalization_45/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %462 = relu(%461.0)
  %463 = pad(%462, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %465 = transpose(%resnet_model/conv2d_46/kernel, axes='(3, 2, 0, 1)')
  %466 = conv2d(%463, %465, kernel_layout='OIHW', use_bias='False', channels='2048', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %471 = batch_norm(%466, %resnet_model/batch_normalization_46/gamma, %resnet_model/batch_normalization_46/beta, %resnet_model/batch_normalization_46/moving_mean, %resnet_model/batch_normalization_46/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %472 = pad(%442, pad_value='0', pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %474 = transpose(%resnet_model/conv2d_43/kernel, axes='(3, 2, 0, 1)')
  %475 = conv2d(%472, %474, kernel_layout='OIHW', use_bias='False', channels='2048', kernel_size='(1, 1)', strides='(2, 2)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %480 = batch_norm(%475, %resnet_model/batch_normalization_43/gamma, %resnet_model/batch_normalization_43/beta, %resnet_model/batch_normalization_43/moving_mean, %resnet_model/batch_normalization_43/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %481 = broadcast_add(%471.0, %480.0)
  %482 = relu(%481)
  %483 = pad(%482, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %485 = transpose(%resnet_model/conv2d_47/kernel, axes='(3, 2, 0, 1)')
  %486 = conv2d(%483, %485, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(1, 1)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %491 = batch_norm(%486, %resnet_model/batch_normalization_47/gamma, %resnet_model/batch_normalization_47/beta, %resnet_model/batch_normalization_47/moving_mean, %resnet_model/batch_normalization_47/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %492 = relu(%491.0)
  %493 = pad(%492, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %495 = transpose(%resnet_model/conv2d_48/kernel, axes='(3, 2, 0, 1)')
  %496 = conv2d(%493, %495, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(3, 3)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %501 = batch_norm(%496, %resnet_model/batch_normalization_48/gamma, %resnet_model/batch_normalization_48/beta, %resnet_model/batch_normalization_48/moving_mean, %resnet_model/batch_normalization_48/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %502 = relu(%501.0)
  %503 = pad(%502, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %505 = transpose(%resnet_model/conv2d_49/kernel, axes='(3, 2, 0, 1)')
  %506 = conv2d(%503, %505, kernel_layout='OIHW', use_bias='False', channels='2048', kernel_size='(1, 1)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %511 = batch_norm(%506, %resnet_model/batch_normalization_49/gamma, %resnet_model/batch_normalization_49/beta, %resnet_model/batch_normalization_49/moving_mean, %resnet_model/batch_normalization_49/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %512 = broadcast_add(%511.0, %482)
  %513 = relu(%512)
  %514 = pad(%513, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %516 = transpose(%resnet_model/conv2d_50/kernel, axes='(3, 2, 0, 1)')
  %517 = conv2d(%514, %516, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(1, 1)', strides='(1, 1)', layout='NCHW', dilation='(1, 1)', padding='[0, 0]')
  %522 = batch_norm(%517, %resnet_model/batch_normalization_50/gamma, %resnet_model/batch_normalization_50/beta, %resnet_model/batch_normalization_50/moving_mean, %resnet_model/batch_normalization_50/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %523 = relu(%522.0)
  %524 = pad(%523, pad_width='((0, 0), (0, 0), (1, 1), (1, 1))')
  %526 = transpose(%resnet_model/conv2d_51/kernel, axes='(3, 2, 0, 1)')
  %527 = conv2d(%524, %526, kernel_layout='OIHW', use_bias='False', channels='512', kernel_size='(3, 3)', strides='(1, 1)', dilation='(1, 1)', padding='[0, 0]', layout='NCHW')
  %532 = batch_norm(%527, %resnet_model/batch_normalization_51/gamma, %resnet_model/batch_normalization_51/beta, %resnet_model/batch_normalization_51/moving_mean, %resnet_model/batch_normalization_51/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %533 = relu(%532.0)
  %534 = pad(%533, pad_width='((0, 0), (0, 0), (0, 0), (0, 0))')
  %536 = transpose(%resnet_model/conv2d_52/kernel, axes='(3, 2, 0, 1)')
  %537 = conv2d(%534, %536, kernel_layout='OIHW', use_bias='False', channels='2048', kernel_size='(1, 1)', strides='(1, 1)', padding='[0, 0]', layout='NCHW', dilation='(1, 1)')
  %542 = batch_norm(%537, %resnet_model/batch_normalization_52/gamma, %resnet_model/batch_normalization_52/beta, %resnet_model/batch_normalization_52/moving_mean, %resnet_model/batch_normalization_52/moving_variance, axis='1', epsilon='1.0009999641624745e-05')
  %543 = broadcast_add(%542.0, %513)
  %544 = relu(%543)
  %545 = mean(%544, axis='(2, 3)', keepdims='True')
  %546 = squeeze(%545, axis='[2, 3]')
  %548 = transpose(%resnet_model/dense/kernel, axes='(1, 0)')
  %549 = dense(%546, %548, units='1001', use_bias='False')
  %551 = broadcast_add(%549, %resnet_model/dense/bias)
  %552 = flatten(%551)
  %553 = softmax(%552)
  ret %553
}
{'resnet_model/transpose/perm': 'int32', 'resnet_model/conv2d/kernel': 'float32', 'resnet_model/batch_normalization/gamma': 'float32', 'resnet_model/batch_normalization/beta': 'float32', 'resnet_model/batch_normalization/moving_mean': 'float32', 'resnet_model/batch_normalization/moving_variance': 'float32', 'resnet_model/conv2d_1/kernel': 'float32', 'resnet_model/batch_normalization_1/gamma': 'float32', 'resnet_model/batch_normalization_1/beta': 'float32', 'resnet_model/batch_normalization_1/moving_mean': 'float32', 'resnet_model/batch_normalization_1/moving_variance': 'float32', 'resnet_model/conv2d_2/kernel': 'float32', 'resnet_model/batch_normalization_2/gamma': 'float32', 'resnet_model/batch_normalization_2/beta': 'float32', 'resnet_model/batch_normalization_2/moving_mean': 'float32', 'resnet_model/batch_normalization_2/moving_variance': 'float32', 'resnet_model/conv2d_3/kernel': 'float32', 'resnet_model/batch_normalization_3/gamma': 'float32', 'resnet_model/batch_normalization_3/beta': 'float32', 'resnet_model/batch_normalization_3/moving_mean': 'float32', 'resnet_model/batch_normalization_3/moving_variance': 'float32', 'resnet_model/conv2d_4/kernel': 'float32', 'resnet_model/batch_normalization_4/gamma': 'float32', 'resnet_model/batch_normalization_4/beta': 'float32', 'resnet_model/batch_normalization_4/moving_mean': 'float32', 'resnet_model/batch_normalization_4/moving_variance': 'float32', 'resnet_model/conv2d_5/kernel': 'float32', 'resnet_model/batch_normalization_5/gamma': 'float32', 'resnet_model/batch_normalization_5/beta': 'float32', 'resnet_model/batch_normalization_5/moving_mean': 'float32', 'resnet_model/batch_normalization_5/moving_variance': 'float32', 'resnet_model/conv2d_6/kernel': 'float32', 'resnet_model/batch_normalization_6/gamma': 'float32', 'resnet_model/batch_normalization_6/beta': 'float32', 'resnet_model/batch_normalization_6/moving_mean': 'float32', 'resnet_model/batch_normalization_6/moving_variance': 'float32', 'resnet_model/conv2d_7/kernel': 'float32', 'resnet_model/batch_normalization_7/gamma': 'float32', 'resnet_model/batch_normalization_7/beta': 'float32', 'resnet_model/batch_normalization_7/moving_mean': 'float32', 'resnet_model/batch_normalization_7/moving_variance': 'float32', 'resnet_model/conv2d_8/kernel': 'float32', 'resnet_model/batch_normalization_8/gamma': 'float32', 'resnet_model/batch_normalization_8/beta': 'float32', 'resnet_model/batch_normalization_8/moving_mean': 'float32', 'resnet_model/batch_normalization_8/moving_variance': 'float32', 'resnet_model/conv2d_9/kernel': 'float32', 'resnet_model/batch_normalization_9/gamma': 'float32', 'resnet_model/batch_normalization_9/beta': 'float32', 'resnet_model/batch_normalization_9/moving_mean': 'float32', 'resnet_model/batch_normalization_9/moving_variance': 'float32', 'resnet_model/conv2d_10/kernel': 'float32', 'resnet_model/batch_normalization_10/gamma': 'float32', 'resnet_model/batch_normalization_10/beta': 'float32', 'resnet_model/batch_normalization_10/moving_mean': 'float32', 'resnet_model/batch_normalization_10/moving_variance': 'float32', 'resnet_model/conv2d_11/kernel': 'float32', 'resnet_model/batch_normalization_11/gamma': 'float32', 'resnet_model/batch_normalization_11/beta': 'float32', 'resnet_model/batch_normalization_11/moving_mean': 'float32', 'resnet_model/batch_normalization_11/moving_variance': 'float32', 'resnet_model/conv2d_12/kernel': 'float32', 'resnet_model/batch_normalization_12/gamma': 'float32', 'resnet_model/batch_normalization_12/beta': 'float32', 'resnet_model/batch_normalization_12/moving_mean': 'float32', 'resnet_model/batch_normalization_12/moving_variance': 'float32', 'resnet_model/conv2d_13/kernel': 'float32', 'resnet_model/batch_normalization_13/gamma': 'float32', 'resnet_model/batch_normalization_13/beta': 'float32', 'resnet_model/batch_normalization_13/moving_mean': 'float32', 'resnet_model/batch_normalization_13/moving_variance': 'float32', 'resnet_model/conv2d_14/kernel': 'float32', 'resnet_model/batch_normalization_14/gamma': 'float32', 'resnet_model/batch_normalization_14/beta': 'float32', 'resnet_model/batch_normalization_14/moving_mean': 'float32', 'resnet_model/batch_normalization_14/moving_variance': 'float32', 'resnet_model/conv2d_15/kernel': 'float32', 'resnet_model/batch_normalization_15/gamma': 'float32', 'resnet_model/batch_normalization_15/beta': 'float32', 'resnet_model/batch_normalization_15/moving_mean': 'float32', 'resnet_model/batch_normalization_15/moving_variance': 'float32', 'resnet_model/conv2d_16/kernel': 'float32', 'resnet_model/batch_normalization_16/gamma': 'float32', 'resnet_model/batch_normalization_16/beta': 'float32', 'resnet_model/batch_normalization_16/moving_mean': 'float32', 'resnet_model/batch_normalization_16/moving_variance': 'float32', 'resnet_model/conv2d_17/kernel': 'float32', 'resnet_model/batch_normalization_17/gamma': 'float32', 'resnet_model/batch_normalization_17/beta': 'float32', 'resnet_model/batch_normalization_17/moving_mean': 'float32', 'resnet_model/batch_normalization_17/moving_variance': 'float32', 'resnet_model/conv2d_18/kernel': 'float32', 'resnet_model/batch_normalization_18/gamma': 'float32', 'resnet_model/batch_normalization_18/beta': 'float32', 'resnet_model/batch_normalization_18/moving_mean': 'float32', 'resnet_model/batch_normalization_18/moving_variance': 'float32', 'resnet_model/conv2d_19/kernel': 'float32', 'resnet_model/batch_normalization_19/gamma': 'float32', 'resnet_model/batch_normalization_19/beta': 'float32', 'resnet_model/batch_normalization_19/moving_mean': 'float32', 'resnet_model/batch_normalization_19/moving_variance': 'float32', 'resnet_model/conv2d_20/kernel': 'float32', 'resnet_model/batch_normalization_20/gamma': 'float32', 'resnet_model/batch_normalization_20/beta': 'float32', 'resnet_model/batch_normalization_20/moving_mean': 'float32', 'resnet_model/batch_normalization_20/moving_variance': 'float32', 'resnet_model/conv2d_21/kernel': 'float32', 'resnet_model/batch_normalization_21/gamma': 'float32', 'resnet_model/batch_normalization_21/beta': 'float32', 'resnet_model/batch_normalization_21/moving_mean': 'float32', 'resnet_model/batch_normalization_21/moving_variance': 'float32', 'resnet_model/conv2d_22/kernel': 'float32', 'resnet_model/batch_normalization_22/gamma': 'float32', 'resnet_model/batch_normalization_22/beta': 'float32', 'resnet_model/batch_normalization_22/moving_mean': 'float32', 'resnet_model/batch_normalization_22/moving_variance': 'float32', 'resnet_model/conv2d_23/kernel': 'float32', 'resnet_model/batch_normalization_23/gamma': 'float32', 'resnet_model/batch_normalization_23/beta': 'float32', 'resnet_model/batch_normalization_23/moving_mean': 'float32', 'resnet_model/batch_normalization_23/moving_variance': 'float32', 'resnet_model/conv2d_24/kernel': 'float32', 'resnet_model/batch_normalization_24/gamma': 'float32', 'resnet_model/batch_normalization_24/beta': 'float32', 'resnet_model/batch_normalization_24/moving_mean': 'float32', 'resnet_model/batch_normalization_24/moving_variance': 'float32', 'resnet_model/conv2d_25/kernel': 'float32', 'resnet_model/batch_normalization_25/gamma': 'float32', 'resnet_model/batch_normalization_25/beta': 'float32', 'resnet_model/batch_normalization_25/moving_mean': 'float32', 'resnet_model/batch_normalization_25/moving_variance': 'float32', 'resnet_model/conv2d_26/kernel': 'float32', 'resnet_model/batch_normalization_26/gamma': 'float32', 'resnet_model/batch_normalization_26/beta': 'float32', 'resnet_model/batch_normalization_26/moving_mean': 'float32', 'resnet_model/batch_normalization_26/moving_variance': 'float32', 'resnet_model/conv2d_27/kernel': 'float32', 'resnet_model/batch_normalization_27/gamma': 'float32', 'resnet_model/batch_normalization_27/beta': 'float32', 'resnet_model/batch_normalization_27/moving_mean': 'float32', 'resnet_model/batch_normalization_27/moving_variance': 'float32', 'resnet_model/conv2d_28/kernel': 'float32', 'resnet_model/batch_normalization_28/gamma': 'float32', 'resnet_model/batch_normalization_28/beta': 'float32', 'resnet_model/batch_normalization_28/moving_mean': 'float32', 'resnet_model/batch_normalization_28/moving_variance': 'float32', 'resnet_model/conv2d_29/kernel': 'float32', 'resnet_model/batch_normalization_29/gamma': 'float32', 'resnet_model/batch_normalization_29/beta': 'float32', 'resnet_model/batch_normalization_29/moving_mean': 'float32', 'resnet_model/batch_normalization_29/moving_variance': 'float32', 'resnet_model/conv2d_30/kernel': 'float32', 'resnet_model/batch_normalization_30/gamma': 'float32', 'resnet_model/batch_normalization_30/beta': 'float32', 'resnet_model/batch_normalization_30/moving_mean': 'float32', 'resnet_model/batch_normalization_30/moving_variance': 'float32', 'resnet_model/conv2d_31/kernel': 'float32', 'resnet_model/batch_normalization_31/gamma': 'float32', 'resnet_model/batch_normalization_31/beta': 'float32', 'resnet_model/batch_normalization_31/moving_mean': 'float32', 'resnet_model/batch_normalization_31/moving_variance': 'float32', 'resnet_model/conv2d_32/kernel': 'float32', 'resnet_model/batch_normalization_32/gamma': 'float32', 'resnet_model/batch_normalization_32/beta': 'float32', 'resnet_model/batch_normalization_32/moving_mean': 'float32', 'resnet_model/batch_normalization_32/moving_variance': 'float32', 'resnet_model/conv2d_33/kernel': 'float32', 'resnet_model/batch_normalization_33/gamma': 'float32', 'resnet_model/batch_normalization_33/beta': 'float32', 'resnet_model/batch_normalization_33/moving_mean': 'float32', 'resnet_model/batch_normalization_33/moving_variance': 'float32', 'resnet_model/conv2d_34/kernel': 'float32', 'resnet_model/batch_normalization_34/gamma': 'float32', 'resnet_model/batch_normalization_34/beta': 'float32', 'resnet_model/batch_normalization_34/moving_mean': 'float32', 'resnet_model/batch_normalization_34/moving_variance': 'float32', 'resnet_model/conv2d_35/kernel': 'float32', 'resnet_model/batch_normalization_35/gamma': 'float32', 'resnet_model/batch_normalization_35/beta': 'float32', 'resnet_model/batch_normalization_35/moving_mean': 'float32', 'resnet_model/batch_normalization_35/moving_variance': 'float32', 'resnet_model/conv2d_36/kernel': 'float32', 'resnet_model/batch_normalization_36/gamma': 'float32', 'resnet_model/batch_normalization_36/beta': 'float32', 'resnet_model/batch_normalization_36/moving_mean': 'float32', 'resnet_model/batch_normalization_36/moving_variance': 'float32', 'resnet_model/conv2d_37/kernel': 'float32', 'resnet_model/batch_normalization_37/gamma': 'float32', 'resnet_model/batch_normalization_37/beta': 'float32', 'resnet_model/batch_normalization_37/moving_mean': 'float32', 'resnet_model/batch_normalization_37/moving_variance': 'float32', 'resnet_model/conv2d_38/kernel': 'float32', 'resnet_model/batch_normalization_38/gamma': 'float32', 'resnet_model/batch_normalization_38/beta': 'float32', 'resnet_model/batch_normalization_38/moving_mean': 'float32', 'resnet_model/batch_normalization_38/moving_variance': 'float32', 'resnet_model/conv2d_39/kernel': 'float32', 'resnet_model/batch_normalization_39/gamma': 'float32', 'resnet_model/batch_normalization_39/beta': 'float32', 'resnet_model/batch_normalization_39/moving_mean': 'float32', 'resnet_model/batch_normalization_39/moving_variance': 'float32', 'resnet_model/conv2d_40/kernel': 'float32', 'resnet_model/batch_normalization_40/gamma': 'float32', 'resnet_model/batch_normalization_40/beta': 'float32', 'resnet_model/batch_normalization_40/moving_mean': 'float32', 'resnet_model/batch_normalization_40/moving_variance': 'float32', 'resnet_model/conv2d_41/kernel': 'float32', 'resnet_model/batch_normalization_41/gamma': 'float32', 'resnet_model/batch_normalization_41/beta': 'float32', 'resnet_model/batch_normalization_41/moving_mean': 'float32', 'resnet_model/batch_normalization_41/moving_variance': 'float32', 'resnet_model/conv2d_42/kernel': 'float32', 'resnet_model/batch_normalization_42/gamma': 'float32', 'resnet_model/batch_normalization_42/beta': 'float32', 'resnet_model/batch_normalization_42/moving_mean': 'float32', 'resnet_model/batch_normalization_42/moving_variance': 'float32', 'resnet_model/conv2d_43/kernel': 'float32', 'resnet_model/batch_normalization_43/gamma': 'float32', 'resnet_model/batch_normalization_43/beta': 'float32', 'resnet_model/batch_normalization_43/moving_mean': 'float32', 'resnet_model/batch_normalization_43/moving_variance': 'float32', 'resnet_model/conv2d_44/kernel': 'float32', 'resnet_model/batch_normalization_44/gamma': 'float32', 'resnet_model/batch_normalization_44/beta': 'float32', 'resnet_model/batch_normalization_44/moving_mean': 'float32', 'resnet_model/batch_normalization_44/moving_variance': 'float32', 'resnet_model/conv2d_45/kernel': 'float32', 'resnet_model/batch_normalization_45/gamma': 'float32', 'resnet_model/batch_normalization_45/beta': 'float32', 'resnet_model/batch_normalization_45/moving_mean': 'float32', 'resnet_model/batch_normalization_45/moving_variance': 'float32', 'resnet_model/conv2d_46/kernel': 'float32', 'resnet_model/batch_normalization_46/gamma': 'float32', 'resnet_model/batch_normalization_46/beta': 'float32', 'resnet_model/batch_normalization_46/moving_mean': 'float32', 'resnet_model/batch_normalization_46/moving_variance': 'float32', 'resnet_model/conv2d_47/kernel': 'float32', 'resnet_model/batch_normalization_47/gamma': 'float32', 'resnet_model/batch_normalization_47/beta': 'float32', 'resnet_model/batch_normalization_47/moving_mean': 'float32', 'resnet_model/batch_normalization_47/moving_variance': 'float32', 'resnet_model/conv2d_48/kernel': 'float32', 'resnet_model/batch_normalization_48/gamma': 'float32', 'resnet_model/batch_normalization_48/beta': 'float32', 'resnet_model/batch_normalization_48/moving_mean': 'float32', 'resnet_model/batch_normalization_48/moving_variance': 'float32', 'resnet_model/conv2d_49/kernel': 'float32', 'resnet_model/batch_normalization_49/gamma': 'float32', 'resnet_model/batch_normalization_49/beta': 'float32', 'resnet_model/batch_normalization_49/moving_mean': 'float32', 'resnet_model/batch_normalization_49/moving_variance': 'float32', 'resnet_model/conv2d_50/kernel': 'float32', 'resnet_model/batch_normalization_50/gamma': 'float32', 'resnet_model/batch_normalization_50/beta': 'float32', 'resnet_model/batch_normalization_50/moving_mean': 'float32', 'resnet_model/batch_normalization_50/moving_variance': 'float32', 'resnet_model/conv2d_51/kernel': 'float32', 'resnet_model/batch_normalization_51/gamma': 'float32', 'resnet_model/batch_normalization_51/beta': 'float32', 'resnet_model/batch_normalization_51/moving_mean': 'float32', 'resnet_model/batch_normalization_51/moving_variance': 'float32', 'resnet_model/conv2d_52/kernel': 'float32', 'resnet_model/batch_normalization_52/gamma': 'float32', 'resnet_model/batch_normalization_52/beta': 'float32', 'resnet_model/batch_normalization_52/moving_mean': 'float32', 'resnet_model/batch_normalization_52/moving_variance': 'float32', 'resnet_model/dense/kernel': 'float32', 'resnet_model/dense/bias': 'float32', 'ArgMax/dimension': 'int32', 'input_tensor': 'float32'}