[Efficientnet/Pytorch] Error when exporting onnx

[aksenventwo]

Related to Model/Framework(s)

[Efficientnet/Pytorch]

Describe the bug

python model2onnx.py --arch efficientnet-quant-b0 --pretrained-from-file nvidia-efficientnet-quant-b0-130421.pth -b 1 --trt True

WARNING: Logging before flag parsing goes to stderr.
E0214 03:47:01.921174 140610225248064 amp_wrapper.py:31] AMP is not avaialble.
=> loading pretrained weights from '/data1/model_zoo/DeepLearningExamples/weights/nvidia-efficientnet-quant-b0-130421.pth'
Traceback (most recent call last):
  File "model2onnx.py", line 161, in <module>
    main(args, model_args, model_arch)
  File "model2onnx.py", line 114, in main
    model = model_arch(**model_args.__dict__)
  File "/data1/model_zoo/DeepLearningExamples/PyTorch/Classification/ConvNets/image_classification/models/model.py", line 149, in __call__
    model.load_state_dict(state_dict)
  File "/data1/env-torch-1.8.x/lib/python3.6/site-packages/torch/nn/modules/module.py", line 1224, in load_state_dict
    self.__class__.__name__, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for EfficientNet:
        Unexpected key(s) in state_dict: "layers.0.block0.se.pooling._input_quantizer._amax", "layers.1.block0.se.pooling._input_quantizer._amax", "layers.1.block1.se.pooling._input_quantizer._amax", "layers.2.block0.se.pooling._input_quantizer._amax", "layers.2.block1.se.pooling._input_quantizer._amax", "layers.3.block0.se.pooling._input_quantizer._amax", "layers.3.block1.se.pooling._input_quantizer._amax", "layers.3.block2.se.pooling._input_quantizer._amax", "layers.4.block0.se.pooling._input_quantizer._amax", "layers.4.block1.se.pooling._input_quantizer._amax", "layers.4.block2.se.pooling._input_quantizer._amax", "layers.5.block0.se.pooling._input_quantizer._amax", "layers.5.block1.se.pooling._input_quantizer._amax", "layers.5.block2.se.pooling._input_quantizer._amax", "layers.5.block3.se.pooling._input_quantizer._amax", "layers.6.block0.se.pooling._input_quantizer._amax", "classifier.pooling._input_quantizer._amax".

and python model2onnx.py --arch efficientnet-b0 --pretrained-from-file nvidia_efficientnet-b0_210412.pth -b 1 --trt True is ok.

To Reproduce Steps to reproduce the behavior:

1. Install '...'
2. Set "..."
3. Launch '...'

Expected behavior A clear and concise description of what you expected to happen.

Environment Please provide at least:

• Container version (e.g. pytorch:19.05-py3):
• GPUs in the system: (e.g. 8x Tesla V100-SXM2-16GB):
• CUDA driver version (e.g. 418.67):

[WuYiYiYi]

Hello, have you solved this problem? I take the same question. @aksenventwo

[aksenventwo]

Hello, have you solved this problem? I take the same question. @aksenventwo

no, i gave up