search

nrimsky / InfluenceFunctions

Implementation of Influence Function approximations for differently sized ML models, using PyTorch
15 stars 2 forks source link

implementation of Proximal Bregman Response Function w.r.t the linear layer of the BERT model #1

Open huutiendang opened 11 months ago

huutiendang commented 11 months ago

Thank you for your implementation. I encountered difficulties while attempting to apply PBRF to the final layer (linear) when fine-tuning a BERT model for text classification. Would it be possible for you to share the code for PBRF in this specific context? @nrimsky

Thank you again!

nrimsky commented 11 months ago

Good idea - I will take a look at this at some point, am currently caught up with other projects but want to get back to Influence Functions stuff shortly

huutiendang commented 11 months ago

Good idea - I will take a look at this at some point, am currently caught up with other projects but want to get back to Influence Functions stuff shortly

Thank you Nina @nrimsky , I hope to hear from you soon!