feat: generator add benchmark mode by setting rule plugins #290
+131
−11
Conversation
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
Signed-off-by: etshen <etshen@nvidia.com>
Ethan-ES
requested review from
a team,
Arsene12358,
jasonqinzhou and
tianhaox
as code owners
simone-chen
reviewed
Jan 28, 2026
Contributor
There was a problem hiding this comment.
Do we want to include cuda_graph_batch_sizes to vllm.rule as well?
Contributor
Author
There was a problem hiding this comment.
Thanks for the review, I've added the cuda_graph_batch_sizes to vllm.rule on line 8 to stay consistent with trtllm and sglang. Please let me know if this is not appropriate.
simone-chen
approved these changes
Jan 28, 2026
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Overview:
To ensure the generator can be used across different environments, we introduced a benchmark mode(running with
--generator-set rule=benchmark) on top of the existing generator. In benchmark mode, the deployment configuration is generated strictly based on the simulation results produced by the aiconfigurator sdk, helping developers align performance more accurately.By default, the generator runs in production mode, which makes trade-offs tailored for general production deployments, such as increasing batch size and reducing the CUDA Graph batch size.
In addition to these two modes, users can define their own rules by adding new folders under
rule_plugin. When invoking the CLI, different rule sets can be selected via the corresponding--generator-set rule=<folder_name>option.