perf: increase Qwen3-235B async rollout GPU memory utilization on H100#3278
Open
humairafirdowse18 wants to merge 1 commit into
Open
perf: increase Qwen3-235B async rollout GPU memory utilization on H100#3278humairafirdowse18 wants to merge 1 commit into
humairafirdowse18 wants to merge 1 commit into
Conversation
Signed-off-by: humairafirdowse18 <humairafirdo@nvidia.com>
humairafirdowse18
added
CI:L1
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Increase
policy.generation.vllm_cfg.gpu_memory_utilizationfrom0.80to0.88for the H100grpo-qwen3-235b-32n8g-async-1offperformance recipe.The additional vLLM memory allocation increases available KV-cache capacity and improves rollout generation throughput, reducing training buffer starvation.
Performance
Averaged over the measured steady-state steps:
0.800.88This is approximately an 11% improvement in E2E throughput.
W&B Logs: https://wandb.ai/nvidia/nemo-rl-perf/runs/lx9ha7b9?nw=nwuserhumairafirdo
Issues
N/A
Usage
No user-facing API changes. The optimization applies automatically when running:
examples/configs/recipes/llm/performance/grpo-qwen3-235b-32n8g-async-1off.yamlBefore your PR is "Ready for review"
Additional Information
This PR contains a single performance-recipe configuration change.