VeRL-Omni v0.3.0: Async Training and Broader Multimodal Post-Training
VeRL-Omni v0.3.0 expands async RL, makes the V1 trainer the default, and adds reproducible recipes for audio-video generation and multimodal understanding.
All posts
Everything published on this blog, newest first. Posts are written by verl maintainers and contributors; if you have shipped something worth explaining, here is how to write it up .
Async RL Training Tuning Guide
Use RL Insight to locate Trainer and Rollouter bubbles, properly balance training and inference resources, maximize single-card throughput, and enable more efficient training.
End-to-End MXFP8 RL on Blackwell: What Matching Buys, When It Pays Off, and How It Fails
MXFP8 on both sides of RL in verl on B200: a sampler on the learner's weight grid cuts train-rollout KL by 16%, training pays off above about 4.5k tokens per micro-batch, and seven failures show what a sync must preserve.
Uni-Agent: Train Long-Horizon Agents at Scale
Uni-Agent is an open framework for scalable long-horizon agent RL. We share our insights on harness adaptation, RL training, and practical recipes for agentic RL.
verl-vla v0.1.0: A Unified Cloud-Edge Post-Training Loop for Large-Scale VLA
verl-vla is an open-source framework built on verl that unifies human-in-the-loop data collection, fine-tuning, evaluation, and reinforcement learning for vision-language-action policies across cloud and edge resources.
Bringing End-to-End LLM Reinforcement Learning to AMD Instinct MI455X with verl
A complete GRPO workflow using verl, Qwen3-30B-A3B, SGLang, and PyTorch FSDP ran for 100 consecutive steps on four AMD Instinct MI455X GPUs, establishing an end-to-end RL baseline for CDNA 5.
VeRL-Omni v0.2.0: Faster Diffusion RL and Stable Omni Training
A release focused on higher-throughput diffusion rollout, reusable omni adapters, and broader recipe coverage.
verl on AMD: production-ready RL post-training on ROCm
RL post-training on AMD Instinct GPUs is here: a turnkey ROCm container, AITER-accelerated vLLM and SGLang rollout, and accuracy validated on both MI300 and MI350 series.
Faster Draft Model Training: verl-SpeCo 0.1.0 Adds Native DSpark Support and Unlocks Standalone Training
verl-SpeCo 0.1.0 adds native DSpark support and standalone draft model training, with reusable feature storage, seven algorithm backends, vLLM/SGLang integration, and GPU/Ascend NPU support.
Introducing verl-tinker: Your Tinker loop, your GPUs
Keep the Tinker Cookbook loop you know, and run SFT, RL, and distillation on verl-managed GPU workers you control.