search

memray / seq2seq-keyphrase

MIT License
318 stars 109 forks source link

Is there a pretrained model available ? #31

Closed pidugusundeep closed 5 years ago

pidugusundeep commented 5 years ago

Hey @memray , I have a newer version of UBUNTU and I cannot really use CUDA 8.0 as it doesn't support ubuntu 18.04, so just wanted to know if you rather have a pre-trained model available to download so that I can use that for testing.

pidugusundeep commented 5 years ago

As described here : https://github.com/memray/seq2seq-keyphrase/blob/cfb2e00bfbae406b87fb09803d76f69bf6f19535/keyphrase/config.py#L435

By any chance, I can get this pre-trained model.

pidugusundeep commented 5 years ago

I was able to download that model and tried to run keyphrase_copynet.py with only setting do_predict = True while I was not able to figure out where exactly the keyphrases are getting generated and stored at the end

memray commented 5 years ago

Hi @pidugusundeep,

I would suggest you use this repo instead. Theano itself is no longer updated and I can run the model after upgrading the Ubuntu. For the PyTorch implementation, for now I don't have any pre-trained model to release. It takes only ~1day to train on KP20k with 1 GPU, so I think training it from scratch won't be a problem.

Best, Rui

pidugusundeep commented 5 years ago

@memray thanks for the clarification, I was able to run this repo and do the predictions but where exactly are the final results stored? is that here? https://github.com/memray/seq2seq-keyphrase/blob/7d989d351d0606106d93ee775ccd63a69ac8fa6b/keyphrase/config.py#L123

If so how do I get the exact keyphrases which are predicted?

memray commented 5 years ago

It should be in the folder as you set. It has been a while and I don't remember the detail now.