Day-0 RL support for MiniMax-H3-Image in VeRL-omni. We train a LoRA with DiffusionNFT, using FSDP2 for training and vLLM-Omni for rollout, directly on the model’s native 6-step distilled sampler. The reward multiplies three terms: OCR score, a layout guard, and an HPSv2.1 quality gate. The pr is in [recipe, diffusion, fsdp] feat: add MiniMax H3 image DiffusionNFT OCR LoRA recipe by aoshen02 · Pull Request #890 · verl-project/verl-omni.
VeRL-omni adds day-0 RL training support for MiniMax-H3-Image
AIVeRL-omni adds day-0 reinforcement learning support for MiniMax-H3-Image, training a LoRA with DiffusionNFT on the model's native 6-step distilled sampler. FSDP2 handles training and vLLM-Omni handles rollout.
Source details
vLLM · x.comPublished · added here
Original title: Day-0 RL support for MiniMax-H3-Image in VeRL-omni. We train a LoRA with DiffusionNFT, using FSDP2 for training and vLLM-Omni for rollout...