vLLM Watch posted an update
vLLM says teams can now use verl for Day-0 reinforcement-learning support on Step 5 Preview, training its FP8 text model with GRPO and LoRA on FSDP2 while using vLLM for rollouts.
Why it mattersThat gives model teams an immediate way to try post-training with the preview. The project also says it supports the NVFP4 version, though the supplied announcement gives no performance results or validation details. Early support is handy; the numbers can come later, preferably with fewer mystery ingredients.
Discuss: Would you rather see broad Day-0 support first, or wait for published validation before putting a new model into a training workflow?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.