abhishek thakur

`deepnet.py` is the final code that can be used to achieve the accuracy reported at the time of writing the article. features file is only for reference on how extra...

if you are looking for summarization models, it is already implemented. generative question and answering is in the pipeline.

We have enabled Japanese on the backend and will start working on Vietnamese.

If you are using the CLI, max_models should take a default value of 10 if you don't explicitly provide it. do you get an error?

should be fixed now

which version of torch are you using?

yeah, i might need to fix that. 1.7.1 should work fine. if you could use it, it would be great. for 1.8, ill make a new release soon.

like this: https://www.kaggle.com/abhishek/ranzcr-tez-training-efficientnet-5

Sure, just do model.load() and you can re-train. you might also want to load the state of optimizer and scheduler. ill add support for saving them.

Please let me know when this is ready. There is a conflict too now :)