From 7b40de99868f9ac56e82b59c036a04e3442bed2e Mon Sep 17 00:00:00 2001 From: hshrivastava-droid Date: Fri, 17 Apr 2026 12:40:59 -0700 Subject: [PATCH 1/4] update vllm flag --- benchmarks/single_node/minimaxm2.5_fp4_b200.sh | 2 ++ perf-changelog.yaml | 6 ++++++ 2 files changed, 8 insertions(+) diff --git a/benchmarks/single_node/minimaxm2.5_fp4_b200.sh b/benchmarks/single_node/minimaxm2.5_fp4_b200.sh index 7cd107a4d..aae5886e1 100755 --- a/benchmarks/single_node/minimaxm2.5_fp4_b200.sh +++ b/benchmarks/single_node/minimaxm2.5_fp4_b200.sh @@ -25,6 +25,8 @@ hf download "$MODEL" SERVER_LOG=/workspace/server.log PORT=${PORT:-8888} +export VLLM_FLOAT32_MATMUL_PRECISION=high + if [ "${DP_ATTENTION}" = "true" ]; then PARALLEL_ARGS="--tensor-parallel-size=1 --data-parallel-size=$TP --enable-expert-parallel" elif [ "$EP_SIZE" -gt 1 ]; then diff --git a/perf-changelog.yaml b/perf-changelog.yaml index f4ba37423..01b01829d 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -1473,3 +1473,9 @@ - "Expand GPT-OSS 120B FP4 MI300X TP=1 concurrency from 64 to 256 for 1k1k" - "Higher concurrency improves MoE weight amortization: 8552 total TPS at conc=256 vs 4016 at conc=64 (2.1x)" pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1053 + +- config-keys: + - minimaxm2.5-fp4-b200-vllm + description: + - "Add VLLM_FLOAT32_MATMUL_PRECISION=high" + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXX From 3a29b80cc17b6f5839f526e16bc0c78da9196d97 Mon Sep 17 00:00:00 2001 From: hshrivastava-droid Date: Fri, 17 Apr 2026 12:46:45 -0700 Subject: [PATCH 2/4] update PR number --- perf-changelog.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/perf-changelog.yaml b/perf-changelog.yaml index 01b01829d..ca63328c6 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -1478,4 +1478,4 @@ - minimaxm2.5-fp4-b200-vllm description: - "Add VLLM_FLOAT32_MATMUL_PRECISION=high" - pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXX + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1069 From 3c10c2b7b2ead35d95d4d0d067f09e9829ed5ed7 Mon Sep 17 00:00:00 2001 From: hshrivastava-droid Date: Fri, 17 Apr 2026 13:27:07 -0700 Subject: [PATCH 3/4] update conc --- .github/configs/nvidia-master.yaml | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-) diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml index b9867c03a..e1b3f4770 100644 --- a/.github/configs/nvidia-master.yaml +++ b/.github/configs/nvidia-master.yaml @@ -3389,21 +3389,20 @@ minimaxm2.5-fp4-b200-vllm: - isl: 1024 osl: 1024 search-space: - - { tp: 1, conc-start: 4, conc-end: 4 } - - { tp: 2, conc-start: 4, conc-end: 512 } - - { tp: 2, ep: 2, conc-start: 128, conc-end: 256 } - - { tp: 2, ep: 2, dp-attn: true, conc-start: 512, conc-end: 512 } - - { tp: 4, conc-start: 4, conc-end: 512 } - - { tp: 4, ep: 4, conc-start: 32, conc-end: 128 } - - { tp: 8, conc-start: 4, conc-end: 4 } + - { tp: 1, conc-start: 4, conc-end: 16 } + - { tp: 2, conc-start: 16, conc-end: 16 } + - { tp: 2, ep: 2, conc-start: 128, conc-end: 128 } + - { tp: 2, ep: 2, dp-attn: true, conc-start: 256, conc-end: 1024 } + - { tp: 4, conc-start: 4, conc-end: 16 } + - { tp: 4, ep: 4, conc-start: 64, conc-end: 128 } + - { tp: 8, conc-start: 4, conc-end: 8 } - isl: 8192 osl: 1024 search-space: - - { tp: 1, conc-start: 4, conc-end: 32 } - - { tp: 1, conc-start: 256, conc-end: 512 } - - { tp: 2, conc-start: 4, conc-end: 512 } + - { tp: 1, conc-start: 256, conc-end: 256 } + - { tp: 1, conc-start: 1024, conc-end: 1024 } - { tp: 2, ep: 2, conc-start: 128, conc-end: 512 } - - { tp: 4, conc-start: 4, conc-end: 512 } + - { tp: 4, conc-start: 4, conc-end: 8 } - { tp: 8, conc-start: 4, conc-end: 4 } # NOTE: At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/MiniMax/MiniMax-M2.html From 5195d6e99dd0fcf01b6453b339089dbf05abeb31 Mon Sep 17 00:00:00 2001 From: hshrivastava-droid Date: Mon, 20 Apr 2026 09:45:52 -0700 Subject: [PATCH 4/4] conc update --- .github/configs/nvidia-master.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml index e1b3f4770..f92fe1d63 100644 --- a/.github/configs/nvidia-master.yaml +++ b/.github/configs/nvidia-master.yaml @@ -3399,8 +3399,8 @@ minimaxm2.5-fp4-b200-vllm: - isl: 8192 osl: 1024 search-space: + - { tp: 1, conc-start: 4, conc-end: 32 } - { tp: 1, conc-start: 256, conc-end: 256 } - - { tp: 1, conc-start: 1024, conc-end: 1024 } - { tp: 2, ep: 2, conc-start: 128, conc-end: 512 } - { tp: 4, conc-start: 4, conc-end: 8 } - { tp: 8, conc-start: 4, conc-end: 4 }