From bfa691ad40dbc3074e00073d9afdaf1f68b37c18 Mon Sep 17 00:00:00 2001 From: Rohit Pujar Nagraj Date: Fri, 5 Jun 2026 16:55:55 -0700 Subject: [PATCH 1/2] feat: expand kimik2.5-fp4-b200-vllm concurrency sweep to 1-128 TP4/EP1 conc 4-64 -> 1-128 and TP8/EP1 conc-start 4 -> 1 for the 1k/1k and 8k/1k cells. Appends a perf-changelog entry. --- .github/configs/nvidia-master.yaml | 8 ++++---- perf-changelog.yaml | 6 ++++++ 2 files changed, 10 insertions(+), 4 deletions(-) diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml index e24ec38c07..e0d26f64c7 100644 --- a/.github/configs/nvidia-master.yaml +++ b/.github/configs/nvidia-master.yaml @@ -2660,13 +2660,13 @@ kimik2.5-fp4-b200-vllm: - isl: 1024 osl: 1024 search-space: - - { tp: 8, ep: 1, conc-start: 4, conc-end: 4 } - - { tp: 4, ep: 1, conc-start: 4, conc-end: 64 } + - { tp: 8, ep: 1, conc-start: 1, conc-end: 4 } + - { tp: 4, ep: 1, conc-start: 1, conc-end: 128 } - isl: 8192 osl: 1024 search-space: - - { tp: 8, ep: 1, conc-start: 4, conc-end: 4 } - - { tp: 4, ep: 1, conc-start: 4, conc-end: 64 } + - { tp: 8, ep: 1, conc-start: 1, conc-end: 4 } + - { tp: 4, ep: 1, conc-start: 1, conc-end: 128 } # NOTE: 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 config reuses the existing diff --git a/perf-changelog.yaml b/perf-changelog.yaml index 47cfcebc12..47339209bf 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -3502,3 +3502,9 @@ - "Update GPT-OSS model for MI355X vLLM from amd/gpt-oss-120b-w-mxfp4-a-fp8 to openai/gpt-oss-120b" pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1670 +- config-keys: + - kimik2.5-fp4-b200-vllm + description: + - "Expand concurrency sweep for the 1k/1k and 8k/1k cells: TP4/EP1 conc 4-64 -> 1-128, TP8/EP1 conc-start 4 -> 1 (conc-end 4 unchanged)." + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/ + From 5bc9600517b5c32cd8c72fc319188b0acbbb3ebc Mon Sep 17 00:00:00 2001 From: Rohit Pujar Nagraj Date: Fri, 5 Jun 2026 16:56:07 -0700 Subject: [PATCH 2/2] chore: set perf-changelog pr-link for kimik2.5-fp4-b200-vllm conc sweep --- perf-changelog.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/perf-changelog.yaml b/perf-changelog.yaml index 47339209bf..68eb1ac368 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -3506,5 +3506,5 @@ - kimik2.5-fp4-b200-vllm description: - "Expand concurrency sweep for the 1k/1k and 8k/1k cells: TP4/EP1 conc 4-64 -> 1-128, TP8/EP1 conc-start 4 -> 1 (conc-end 4 unchanged)." - pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/ + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1677