Is there a way to do less computation for T5 and DPR models?

weaviate / t2v-transformers-models

This is the repo for the container that holds the models for the text2vec-transformers module

BSD 3-Clause "New" or "Revised" License

39 stars 27 forks source link

Is there a way to do less computation for T5 and DPR models? #63

Open tiejunhu opened 1 year ago

tiejunhu commented 1 year ago

After reading the vectorizer.py, I found that the models do all the computation but we only take the output from a middle(hidden) layer in T5Model and DPRModel. Is there a way to prevent the models from doing all the work after the hidden layer?