Add Kimi K2.5 1T INT4 vLLM benchmark for B200#735
Open
functionstackx wants to merge 12 commits intomainfrom
Open
Add Kimi K2.5 1T INT4 vLLM benchmark for B200#735functionstackx wants to merge 12 commits intomainfrom
functionstackx wants to merge 12 commits intomainfrom
Conversation
- Add kimik2.5-int4-mi355x-vllm config to amd-master.yaml - Image: vllm/vllm-openai-rocm:v0.15.1 (per Andy Luo's recipe) - Model: moonshotai/Kimi-K2.5 with --mm-encoder-tp-mode data - TP=8, concurrency 4-64 for 1k1k, 1k8k, 8k1k - No AITER env vars, no --no-enable-prefix-caching Closes #723 Co-authored-by: functionstackx <functionstackx@users.noreply.github.com>
Co-authored-by: functionstackx <functionstackx@users.noreply.github.com>
The flag is not recognized by vllm v0.15.1 and causes the server to fail on startup. Co-authored-by: functionstackx <functionstackx@users.noreply.github.com>
The benchmark_serving.py script already accepts --trust-remote-code but benchmark_lib.sh's run_benchmark_serving() function wasn't passing it through. This caused tokenizer loading failures for models like Kimi-K2.5 that require trust_remote_code=True. - Add --trust-remote-code flag parsing in run_benchmark_serving() - Pass the flag through to benchmark_serving.py when set - Enable --trust-remote-code in the Kimi-K2.5 benchmark script Co-authored-by: functionstackx <functionstackx@users.noreply.github.com>
- Create benchmarks/kimik2.5_int4_b200.sh with --trust-remote-code on both vllm serve and run_benchmark_serving - Add kimik2.5-int4-b200-vllm config to nvidia-master.yaml - Update perf-changelog.yaml with new entry Image: vllm/vllm-openai:v0.15.1 Model: moonshotai/Kimi-K2.5, TP=8, concurrency 4-64 Flags: --mm-encoder-tp-mode data, --trust-remote-code Closes #727 Co-authored-by: functionstackx <functionstackx@users.noreply.github.com>
Contributor
Author
|
/sweep |
Contributor
|
@functionstackx Kicking off a sweep. Run: https://github.com/SemiAnalysisAI/InferenceX/actions/runs/22127197803 |
Contributor
Author
|
/sweep test-config --config-keys kimik2.5-int4-b200-vllm --runner-config .github/configs/runners.yaml --config-files .github/configs/nvidia-master.yaml |
functionstackx
changed the base branch from
claude/issue-723-20260218-0123
to
main
functionstackx
changed the title
Collaborator
|
I will take this over |
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
2 participants
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.
Summary
Closes #727
Generated with Claude Code