Flair: Google Universal Sentence Encoder

[mvss80]

Is it possible to add Google's Transformer Model available at https://tfhub.dev/google/universal-sentence-encoder-large/3 as part of Document Embeddings within Flair?

I couldn't find any pytorch port of that model and it would be very handy to be able to run comparisons.

[alanakbik]

Yes, that is a great idea. We would like to get more transformer architectures into Flair, for instance to train transformer-based LMs.

I am not sure when we can get around to implementing this, so any help from the community would be greatly appreciated!

[stefan-it]

I think InferSent[1] would be a good alternative and a PyTorch implementation exists :)

---

[1] [Supervised Learning of Universal Sentence Representations from Natural Language Inference Data](https://arxiv.org/abs/1705.02364)

[franz101]

https://github.com/helloeve/universal-sentence-encoder-fine-tune/blob/master/universal_sentence_encoder_fine_tune.ipynb

Here is also a good example. Who would like to implement it with me?

[franz101]

Here is also the link for the pertained large english model: https://storage.googleapis.com/tfhub-modules/google/universal-sentence-encoder-large/3.tar.gz

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.