Documentation for IndicGLUE

[ameet-1997]

Hi,

This repository has some great resources!

I was wondering where the documentation for the fine-tuning datasets are. Sorry if I overlooked something obvious. I want to evaluate my models (not IndicBERT) on your dataset, so wanted to see how that is possible.

Thanks!

[anoopkunchukuttan]

Hi Ameet,

You can find more details here: https://indicnlp.ai4bharat.org/

Specifically about the datasets here: https://indicnlp.ai4bharat.org/indic-glue/

And paper about this work here: https://indicnlp.ai4bharat.org/papers/arxiv2020_indicnlp_corpus.pdf

[ameet-1997]

Thanks for the reply! I've gone through all these resources.

However, I was referring to documentation on the code for fine-tuning external models. Is this notebook (link) the only available resource for that? Or is there a place I can look for the documentation?

[divkakwani]

Hey Ameet. Unfortunately, we don't have a documentation yet. We are working on adding both user and developer documentation in our repository.

For now, depending on your use case, you can check if the following helps:

• To get all the parameters, use python3 -m fine_tune.cli --help
• To check all the available datasets: Refer the ALL_DATASETS dictionary in https://github.com/AI4Bharat/indic-bert/blob/master/fine_tune/cli.py and use the corresponding key in the colab notebook.
• Also note that you can use any Huggingface model with the code, you just have to pass its huggingface shortcut name as model parameter. (e.g. for mbert put bert-base-multilingual-cased) For list of all available names, please refer https://huggingface.co/transformers/pretrained_models.html

It is also possible to run the evaluation on non-huggingface models. I will shortly put a documentation on it and update here.

[ameet-1997]

Thanks so much! Leaving this open so that you can close it after you have updated the documentation, but feel free to close it!

[divkakwani]

Hey, I have added some usage notes here https://github.com/AI4Bharat/indic-bert/blob/master/docs/advanced-usage.md. Does it answer your query?

Also, feel free to add here anything else that you'd want to see in the documentation.