How to train a multi-agent video understanding task

Open 64327069 opened this issue 3 months ago • 6 comments

I have read your paper, which have great help for my work. In your paper, you said that agent-lightning supports multi-agent reinforcement learning with any agent and almost no code change. However, I can not read how to train a multi-agent (eg. 2 Qwen2.5-VL-7B training at the same time）system in the readme and guidance. Can you add the content about how to train a multi-agent framework in video-understanding?

Sep 24 '25 03:09 64327069

We can't train two models at the same time due to a verl limitation.

We can however train two agents alternatively by specifying the trained_agents parameter in LitAgent.

Sep 26 '25 05:09 ultmaster

Do you have any examples or guidance about training 2 agents alternatively?

Sep 29 '25 03:09 64327069