Skip to content
Trending storyDeveloping

VeRL-omni adds day-0 reinforcement learning support for MiniMax-H3-Image

1 article1 sourcesince Nov 10Last article 1h ago ·

Overview

AISummary of 1 article

VeRL-omni has added day-0 reinforcement learning support for MiniMax-H3-Image, according to a post from the vLLM project on X. The recipe trains a LoRA with DiffusionNFT on the model's native 6-step distilled sampler, uses FSDP2 for training and vLLM-Omni for rollout, and is available in pull request #890 on verl-project/verl-omni.

The reward multiplies three signals: an OCR score, a layout guard, and an HPSv2.1 quality gate.

Written by AI from the articles below · updated Nov 10, 10:54 AM ET

Check the sources:

Latest developmentNov 10, 10:43 AM ET
VeRL-omni adds day-0 RL training support for MiniMax-H3-Image

Article timeline

The articles in this story. Times are ET.

Nov 10
  1. vLLMOfficial
    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. The reward multiplies an OCR score, a layout guard, and an HPSv2.1 quality gate, and the recipe is in pull request #890 on verl-project/verl-omni.

    Image from @vllm_project's post

Heat Trend

Not enough continuous observations to show a trend yet.