search

BlackSamorez / tensor_parallel

Automatically split your PyTorch models on multiple GPUs for training & inference
MIT License
629 stars 39 forks source link

LLaMA-2 #101

Closed BlackSamorez closed 1 year ago

BlackSamorez commented 1 year ago

The latest transformers require accelerate>=0.20.3. accelerate==0.15.0 is used here for testing purposes. It needs to be updated. edit: latest transformers also changed LLaMA modeling so this is now a LLaMA-2 PR.

BlackSamorez commented 1 year ago

I also freezed peft==0.3.0 version for consistency.