You have put in some great work! :) We recently updated the package for PyTorch v1.3, if there are certain features that you think may be useful to all users of deepmatcher, it'll be great if you could create more modular pull requests for specific components, I'll be happy to merge them to master.

You have put in some great work! :) We recently updated the package for PyTorch v1.3, if there are certain features that you think may be useful to all users of deepmatcher, it'll be great if you could create more modular pull requests for specific components, I'll be happy to merge them to master.

Thank you very much!
I really think that a big improvement would be to allow users to load a dataset directly as a Pandas.DataFrame instead of loading it from csv files. I've already done some work but it's still incomplete and it's a kind of hard fix.
Another useful improvement would be the possibility to load a self-pretrained embeddings; as for the above suggestion, I've done something, but it's a hard fix