search

erfanzar / EasyDeL

Accelerate, Optimize performance with streamlined training and serving options with JAX.
https://easydel.readthedocs.io/en/latest/
Apache License 2.0
208 stars 25 forks source link

How to do sequence classification training ? #169

Closed sparsh35 closed 1 month ago

sparsh35 commented 2 months ago

Describe the bug I want to train a reward model using Easydel with sequence classification. The classifier has been implemented in the Flax sequence classifier classes for each model, but is there any way to load a model to directly with sequence classifier and train it To Reproduce Steps to reproduce the behavior

erfanzar commented 2 months ago

hi that's not supported at the moment, but i guess ill support that in next 24 hour

sparsh35 commented 1 month ago

I am also thinking of implementing Online DPO trainer with EasyDel, with a little bit of your support if you are interested. It can be comparable with PPO as per paper from Deepmind, the biggest bottleneck I guess would be in generation of completitions during training.

erfanzar commented 1 month ago

Yes seems cool to me If u needed any helps u can dm me in discord

erfanzar commented 1 month ago

@sparsh35 SequenceClassificationTrainer is added