verl
verl copied to clipboard
Question about multi image input support.
Thanks for open source. Does verl support inputting multiple images into VLM (such as Qwen2.5VL) for GRPO based RL?