Section under construction. Feel free to contribute!
Online methods such as GRPO or Online DPO require the model to generate completions, which is often a slow process and can significantly impact training time. To speed up generation, you can use vLLM, a library that enables fast generation through, among other things, PagedAttention. TRL’s online trainers support vLLM, greatly improving training speed.
To use vLLM, first install it using:
pip install vllm
or
pip install "trl[vllm]"
Then, enable it by passing use_vllm=True
in the training arguments.
from trl import OnlineDPOConfig
training_args = OnlineDPOConfig(..., use_vllm=True)