Build TF pre-trained model in Pytorch


I have a pre-trained model implemented and trained in TensorFlow and saved into the three common TF files (’.data’ , ‘.index’ and ‘.meta’).

Is there any way for trans-coding these files into another format that Pytorch can load and directly build the same trained model accordingly ?

I have heard about ‘onnex’ but still don’t know if it is relevant to such task.

Best Regards