Convert Special ResNet101 from TensorFlow to PyTorch

[ahkarami]

Dear @kitstar,
By using your complete tutorial I can convert ResNet V1 101 model (from TF-Slim) to PyTorch, and the converted model works correct. Now I want to convert another ResNet V1 101 model which has 5,000 classes from TensorFlow to PyTorch. In fact, I want to convert the pre-trained Open Images ResNet V1 101 from TensorFlow to PyTorch.
To this end, I have tested with follow steps:

1. Download resnet_v1_101 data:

$ wget https://storage.googleapis.com/openimages/2017_07/oidv2-resnet_v1_101.ckpt.tar.gz
$ tar -xvf oidv2-resnet_v1_101.ckpt.tar.gz

2. Extract tf model files:
   python -m mmdnn.conversion.examples.tensorflow.extract_model -n resnet_v1_101 -ckpt oidv2-resnet_v1_101.ckpt.data-00000-of-00001

However, I got a long error like this:

... DataLossError (see above for traceback): Unable to open table file ./oidv2-resnet_v1_101.ckpt.data-00000-of-00001: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?
	 [[Node: save/RestoreV2 = RestoreV2[dtypes=[DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_save/Const_0_0, save/RestoreV2/tensor_names, save/RestoreV2/shape_and_slices)]]

I used the TF 1.4.1 and PyTorch 0.3.
Would you please help me to address this issue and do the conversion?

[kitstar]

Hi, the extract_model.py of tf is an example code to show how to generate the full checkpoint files. You should modify your own inference code to generate your own checkpoint files.