From 8babce4f623634142418e4a79aad36f35843dc48 Mon Sep 17 00:00:00 2001 From: hshrivastava-droid Date: Mon, 20 Apr 2026 18:53:17 -0700 Subject: [PATCH 1/3] update vllm --- .github/configs/nvidia-master.yaml | 11 ++++++----- benchmarks/single_node/minimaxm2.5_fp8_b300.sh | 2 +- perf-changelog.yaml | 6 ++++++ 3 files changed, 13 insertions(+), 6 deletions(-) diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml index d6202608d..1703b2a33 100644 --- a/.github/configs/nvidia-master.yaml +++ b/.github/configs/nvidia-master.yaml @@ -3538,15 +3538,16 @@ minimaxm2.5-fp8-b300-vllm: - isl: 1024 osl: 1024 search-space: - - { tp: 2, conc-start: 4, conc-end: 512 } - - { tp: 4, conc-start: 4, conc-end: 512 } - - { tp: 2, ep: 2, conc-start: 512, conc-end: 512 } + - { tp: 4, conc-start: 4, conc-end: 128 } - { tp: 4, ep: 4, conc-start: 256, conc-end: 512 } + - { tp: 2, ep: 2, conc-start: 512, conc-end: 1024 } + - { tp: 2, ep: 2, dp-attn: true, conc-start: 1024, conc-end: 2048 } - isl: 8192 osl: 1024 search-space: - - { tp: 2, conc-start: 4, conc-end: 512 } - - { tp: 4, conc-start: 4, conc-end: 512 } + - { tp: 1, conc-start: 4, conc-end: 16 } + - { tp: 2, conc-start: 64, conc-end: 256 } + - { tp: 4, conc-start: 4, conc-end: 8 } minimaxm2.5-fp4-b200-vllm: image: vllm/vllm-openai:v0.19.0-cu130 diff --git a/benchmarks/single_node/minimaxm2.5_fp8_b300.sh b/benchmarks/single_node/minimaxm2.5_fp8_b300.sh index 210109e89..3e83a0b36 100755 --- a/benchmarks/single_node/minimaxm2.5_fp8_b300.sh +++ b/benchmarks/single_node/minimaxm2.5_fp8_b300.sh @@ -28,7 +28,7 @@ hf download "$MODEL" SERVER_LOG=/workspace/server.log PORT=${PORT:-8888} -export VLLM_FLASHINFER_ALLREDUCE_BACKEND=mnnvl +export VLLM_FLOAT32_MATMUL_PRECISION=high if [ "$EP_SIZE" -gt 1 ]; then EP=" --enable-expert-parallel" diff --git a/perf-changelog.yaml b/perf-changelog.yaml index 75986a0eb..b8b136bd8 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -1646,3 +1646,9 @@ description: - "Add kv-cache-dtype fp8, max-cudagraph-capture-size 2048, max-num-batched-tokens, and stream-interval 20 to server launch args" pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1047 + +- config-keys: + - minimaxm2.5-fp8-b300-vllm + description: + - "Add VLLM_FLOAT32_MATMUL_PRECISION=high, remove VLLM_FLASHINFER_ALLREDUCE_BACKEND=mnnvl" + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXX From b256571c507e60a1a5547bce82ee4a27e49f1004 Mon Sep 17 00:00:00 2001 From: hshrivastava-droid Date: Mon, 20 Apr 2026 19:00:27 -0700 Subject: [PATCH 2/3] 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 b8b136bd8..5ad0f32d4 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -1651,4 +1651,4 @@ - minimaxm2.5-fp8-b300-vllm description: - "Add VLLM_FLOAT32_MATMUL_PRECISION=high, remove VLLM_FLASHINFER_ALLREDUCE_BACKEND=mnnvl" - pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXX + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1106 From 54bf90ee3c3c72d0df035bc310b9785508028df6 Mon Sep 17 00:00:00 2001 From: hshrivastava-droid Date: Tue, 21 Apr 2026 14:25:59 -0700 Subject: [PATCH 3/3] update conc --- .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 1703b2a33..7c4fcab9a 100644 --- a/.github/configs/nvidia-master.yaml +++ b/.github/configs/nvidia-master.yaml @@ -3541,7 +3541,7 @@ minimaxm2.5-fp8-b300-vllm: - { tp: 4, conc-start: 4, conc-end: 128 } - { tp: 4, ep: 4, conc-start: 256, conc-end: 512 } - { tp: 2, ep: 2, conc-start: 512, conc-end: 1024 } - - { tp: 2, ep: 2, dp-attn: true, conc-start: 1024, conc-end: 2048 } + - { tp: 2, ep: 2, dp-attn: true, conc-start: 1024, conc-end: 1024 } - isl: 8192 osl: 1024 search-space: