the example model of 'SGPT-125M-weightedmean-nli-bitfit' not compatible with ST

Muennighoff / sgpt

SGPT: GPT Sentence Embeddings for Semantic Search

https://arxiv.org/abs/2202.08904

MIT License

841 stars 51 forks source link

the example model of 'SGPT-125M-weightedmean-nli-bitfit' not compatible with ST #29

Open dengwx2009 opened 1 year ago

dengwx2009 commented 1 year ago

hi, I tried to run the example be_st with 'Muennighoff/SGPT-125M-weightedmean-nli-bitfit' but it throw out the exception as follows: /usr/local/lib/python3.8/site-packages/sentence_transformers/models/Pooling.py:120 in load TypeError: __init__() got an unexpected keyword argument 'pooling_mode_weightedmean_tokens' Is ’pooling_mode_weightedmean_tokens‘ in the config.json changed to ’pooling_mode_mean_tokens’ ？

Muennighoff commented 1 year ago

Can you try what is explained here: https://github.com/Muennighoff/sgpt/issues/14#issuecomment-1352628910 You're likely using an old ST version

MattiL commented 1 year ago

I got the same error with the computing of the web interface at https://huggingface.co/bigscience-data/sgpt-bloom-1b7-nli a moment ago.

Muennighoff commented 1 year ago

I got the same error with the computing of the web interface at https://huggingface.co/bigscience-data/sgpt-bloom-1b7-nli a moment ago.

Yes that's because the web interface is not using the latest ST version