From 59401ac9f18c246dceffce28bc878881179405e4 Mon Sep 17 00:00:00 2001 From: Qiuyang Yue Date: Tue, 9 Jun 2026 04:15:51 -0700 Subject: [PATCH] [Kernel][Perf] Tune fused_moe FP8 config for Qwen3-Next-80B tp=4 on H100 (+25% at batch 96-512) (#44830) Signed-off-by: Qiuyang Yue --- ...,dtype=fp8_w8a8,block_shape=[128,128].json | 43 +++++++++++++++++++ 1 file changed, 43 insertions(+) create mode 100644 vllm/model_executor/layers/fused_moe/configs/E=512,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128,128].json diff --git a/vllm/model_executor/layers/fused_moe/configs/E=512,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128,128].json b/vllm/model_executor/layers/fused_moe/configs/E=512,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128,128].json new file mode 100644 index 0000000000..fc22e54982 --- /dev/null +++ b/vllm/model_executor/layers/fused_moe/configs/E=512,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128,128].json @@ -0,0 +1,43 @@ +{ + "triton_version": "3.6.0", + "96": { + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "BLOCK_SIZE_K": 128, + "GROUP_SIZE_M": 64, + "num_warps": 4, + "num_stages": 2 + }, + "128": { + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "BLOCK_SIZE_K": 128, + "GROUP_SIZE_M": 64, + "num_warps": 4, + "num_stages": 2 + }, + "256": { + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "BLOCK_SIZE_K": 128, + "GROUP_SIZE_M": 64, + "num_warps": 4, + "num_stages": 2 + }, + "512": { + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "BLOCK_SIZE_K": 256, + "GROUP_SIZE_M": 64, + "num_warps": 4, + "num_stages": 2 + }, + "1024": { + "BLOCK_SIZE_M": 32, + "BLOCK_SIZE_N": 128, + "BLOCK_SIZE_K": 128, + "GROUP_SIZE_M": 64, + "num_warps": 4, + "num_stages": 2 + } +} \ No newline at end of file