Question

可以从pycaffe读取caffe模型的BathNorm和Scale重量，这是BatchNorm中的三个权重和Scale中的两个权重。我尝试使用以下代码将这些砝码复制到pytorch BatchNorm：

if 'conv3_final_bn' == name:
    assert len(blobs) == 3, '{} layer blob count: {}'.format(name, len(blobs))
    torch_mod['conv3_final_bn.running_mean'] = blobs[0].data
    torch_mod['conv3_final_bn.running_var'] = blobs[1].data
elif 'conv3_final_scale' == name:
    assert len(blobs) == 2, '{} layer blob count: {}'.format(name, len(blobs))
    torch_mod['conv3_final_bn.weight'] = blobs[0].data
    torch_mod['conv3_final_bn.bias'] = blobs[1].data

两个BatchNorm的行为不同。我还尝试设置conv3_final_bn.weight = 1和conv3_final_bn.bias = 0来验证caffe的BN层，结果也不匹配。

我应该如何处理错误的匹配？

Answer 1

知道了！ caffe的BatchNorm中仍有第三个参数。代码应为：

if 'conv3_final_bn' == name:
    assert len(blobs) == 3, '{} layer blob count: {}'.format(name, len(blobs))
    torch_mod['conv3_final_bn.running_mean'] = blobs[0].data / blobs[2].data[0]
    torch_mod['conv3_final_bn.running_var'] = blobs[1].data / blobs[2].data[0]
elif 'conv3_final_scale' == name:
    assert len(blobs) == 2, '{} layer blob count: {}'.format(name, len(blobs))
    torch_mod['conv3_final_bn.weight'] = blobs[0].data
    torch_mod['conv3_final_bn.bias'] = blobs[1].data

怎么把Caffe的BatchNorm重量转换成pytorch BathNorm？

1 个答案: