search

EmilyAlsentzer / clinicalBERT

repository for Publicly Available Clinical BERT Embeddings
MIT License
673 stars 135 forks source link

Interagtion with Huggingface/Transformers #19

Closed NeelKanwal closed 4 years ago

NeelKanwal commented 4 years ago

Hi,

Can you please add this into huggingface/Transformer community model. It can be very useful to avail of built-in functions with the transformer library.

https://github.com/huggingface/transformers/tree/master/templates/adding_a_new_example_script

Here are the details. It will be really helpful to test it on existing scripts.

Thanks

Kanwal

julien-c commented 4 years ago

That would be awesome! (and just uploading the weights should be enough given that https://github.com/EmilyAlsentzer/clinicalBERT/issues/3#issuecomment-492793549)

EmilyAlsentzer commented 4 years ago

You can now find the clinicalBERT models in the Transformers model hub. Check out the Bio+Clinical BERT and Bio+Discharge Summary BERT model pages for instructions on how to use the models within the Transformers library. Model cards will be added to the library shortly (pending an approved pull request. Stay tuned for an example on how to run Med-NLI using the library.

Hope this helps improve usability of the models! Feel free to reopen the issue if you have other feature requests.

harpap commented 2 years ago

Hello @EmilyAlsentzer, Thank you for this repo.

Is it possible to also add the i2b2 2010 on HuggingFace? I am having a hard time trying to run it. Thanks in advance

tnaumann commented 2 years ago

@harpap - in general, it is not possible to upload i2b2 datasets to HuggingFace because users of i2b2 datasets must register AND submit a DUA for access (see https://www.i2b2.org/NLP/DataSets/Main.php); uploading the dataset to HuggingFace would circumvent these requirements.

However, I would greatly encourage you to connect with the BigScience Workshop initiative since they have been creating a number of dataset loaders for biomedical datasets: https://github.com/orgs/bigscience-workshop/projects/6/views/1.

HuggingFace has also recently begun to support gated access to datasets. While i2b2 has not uploaded datasets yet, having appropriate dataset loaders ready to go could be the nudge needed!

julien-c commented 2 years ago

@huggingface CTO here – if there are any features we can build in to "gated datasets" to make them suitable for your datasets' workflows, let me know! we'd be happy to build them