Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
sergiopaniego 
posted an update Sep 22
Post
458
ICYMI, Async GRPO in TRL now supports LoRA and we wrote a looong blog testing it

> the adapter is a few megabytes, so the weight sync is a file instead of an NCCL transfer
> 3 HF Jobs: 1 trainer and 2 vLLM replicas
> the adapter travels through an HF Storage Bucket mounted in all 3 at the same path
> a proxy in front of the replicas routes each rollout to the one already holding its KV prefix

https://huggingface.co/blog/asyncgrpo-lora-hfjobs
In this post