How to save model for Neural Machine Translation ?

[wannaphong]

I want to save model for Neural Machine Translation (https://nbviewer.jupyter.org/github/DSKSD/DeepNLP-models-Pytorch/blob/master/notebooks/07.Neural-Machine-Translation-with-Attention.ipynb). Can you help me ?

[simonjisu]

you can find how to save your model in pytorch documentation. SAVING AND LOADING MODELS

In that notebook, insert theses code after model trained.

torch.save({
            'encoder_state_dict': encoder.state_dict(),
            'decoder_state_dict': decoder.state_dict(),
            }, YOUR_PATH)

When loading model, run following codes

encoder = Encoder(...)   # set same parameters as you trained
decoder = Decoder(...)  # set same parameters as you trained
checkpoint = torch.load(YOUR_PATH)
encoder.load_state_dict(checkpoint['encoder_state_dict'])
decoder.load_state_dict(checkpoint['decoder_state_dict'])

or, you can save 2 paths to your model

torch.save(encoder.state_dict(), YOUR_ENCODER_PATH)
torch.save(decoder.state_dict(), YOUR_DECODER_PATH)

When loading model, don't forget load both paths.

encoder = Encoder(...)   # set same parameters as you trained
decoder = Decoder(...)  # set same parameters as you trained
encoder.load_state_dict(torch.load(YOUR_ENCODER_PATH))
decoder.load_state_dict(torch.load(YOUR_DECODER_PATH))