diff --git a/perf-changelog.yaml b/perf-changelog.yaml index 2d37ff5d4..e25542834 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -2131,3 +2131,12 @@ - "Search-space: tp=8 and tp=4/ep=1 over conc 4-64, on both 1024/1024 and 8192/1024 ISL/OSL" - "At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/moonshotai/Kimi-K2.5.html does not have a B300-specific recipe, so this reuses the existing Kimi-K2.5 INT4 B200 vLLM recipe as-is until B300-specific tuning is available" pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1057 + +- config-keys: + - kimik2.5-int4-b300-vllm + description: + - "Add Kimi-K2.5 INT4 B300 vLLM benchmark" + - "Image: vllm/vllm-openai:v0.20.0-cu130" + - "Search-space: tp=8 and tp=4/ep=1 over conc 4-64, on both 1024/1024 and 8192/1024 ISL/OSL" + - "At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/moonshotai/Kimi-K2.5.html does not have a B300-specific recipe, so this reuses the existing Kimi-K2.5 INT4 B200 vLLM recipe as-is until B300-specific tuning is available" + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1057