Yuge Zhang
Yuge Zhang
You can ignore the failure when vLLM stream=True. Addressing that is a separate task that is being processed in parallel. Could you explain how you were going to run the...
It's already in the roadmap.
@flatLying I think you are talking about a different thing. Do you mean an agent without chat.completion and calling HuggingFace transformer generate function directly?
I don't see what's the difference between dynamic workflows and static workflows in case of RL training?
They are the same for our framework. We can write an example but we don't have enough manpowers yet.
Agent-lightning currently has a workaround to strip the streaming mode into non-streaming mode via LiteLLM proxy: #293 The LiteLLM telemetry also has some severe bugs relating to streaming. So vLLM...
Backlog v0.3 - what's on my mind: - Tinker support - Azure OpenAI SFT support (cloud-sft branch) - SqliteLightningStore - Hao's improvement on tracer - Online RL example - VERL...