From 5b9f3080ab87ffc79a79a1972d03b71bb6e218bb Mon Sep 17 00:00:00 2001 From: sufubao <1411045825@qq.com> Date: Wed, 3 Sep 2025 15:02:27 +0800 Subject: [PATCH] Qwen configs --- ...num=1,use_fp8_w8a8=false}_NVIDIA_H200.json | 110 ++++++++++++++++++ ...num=8,use_fp8_w8a8=false}_NVIDIA_H200.json | 110 ++++++++++++++++++ 2 files changed, 220 insertions(+) create mode 100644 lightllm/common/triton_utils/autotune_kernel_configs/triton_3.4.0/NVIDIA_H200/grouped_matmul:v1/{K=192,N=4096,expert_num=128,mul_routed_weight=true,out_dtype=torch.bfloat16,topk_num=1,use_fp8_w8a8=false}_NVIDIA_H200.json create mode 100644 lightllm/common/triton_utils/autotune_kernel_configs/triton_3.4.0/NVIDIA_H200/grouped_matmul:v1/{K=4096,N=384,expert_num=128,mul_routed_weight=false,out_dtype=torch.bfloat16,topk_num=8,use_fp8_w8a8=false}_NVIDIA_H200.json diff --git a/lightllm/common/triton_utils/autotune_kernel_configs/triton_3.4.0/NVIDIA_H200/grouped_matmul:v1/{K=192,N=4096,expert_num=128,mul_routed_weight=true,out_dtype=torch.bfloat16,topk_num=1,use_fp8_w8a8=false}_NVIDIA_H200.json b/lightllm/common/triton_utils/autotune_kernel_configs/triton_3.4.0/NVIDIA_H200/grouped_matmul:v1/{K=192,N=4096,expert_num=128,mul_routed_weight=true,out_dtype=torch.bfloat16,topk_num=1,use_fp8_w8a8=false}_NVIDIA_H200.json new file mode 100644 index 000000000..e16155639 --- /dev/null +++ b/lightllm/common/triton_utils/autotune_kernel_configs/triton_3.4.0/NVIDIA_H200/grouped_matmul:v1/{K=192,N=4096,expert_num=128,mul_routed_weight=true,out_dtype=torch.bfloat16,topk_num=1,use_fp8_w8a8=false}_NVIDIA_H200.json @@ -0,0 +1,110 @@ +{ + "1024": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 64, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "128": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "131072": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 128, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 16, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "16384": { + "BLOCK_SIZE_K": 32, + "BLOCK_SIZE_M": 64, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 4, + "num_warps": 4 + }, + "2048": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 32, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 64, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "256": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 2, + "num_warps": 4 + }, + "32768": { + "BLOCK_SIZE_K": 32, + "BLOCK_SIZE_M": 64, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 32, + "NEED_TRANS": false, + "num_stages": 4, + "num_warps": 4 + }, + "512": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "64": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 64, + "GROUP_SIZE_M": 64, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "8": { + "BLOCK_SIZE_K": 32, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "800": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "8192": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 64, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 16, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + } + } \ No newline at end of file diff --git a/lightllm/common/triton_utils/autotune_kernel_configs/triton_3.4.0/NVIDIA_H200/grouped_matmul:v1/{K=4096,N=384,expert_num=128,mul_routed_weight=false,out_dtype=torch.bfloat16,topk_num=8,use_fp8_w8a8=false}_NVIDIA_H200.json b/lightllm/common/triton_utils/autotune_kernel_configs/triton_3.4.0/NVIDIA_H200/grouped_matmul:v1/{K=4096,N=384,expert_num=128,mul_routed_weight=false,out_dtype=torch.bfloat16,topk_num=8,use_fp8_w8a8=false}_NVIDIA_H200.json new file mode 100644 index 000000000..2488ede5d --- /dev/null +++ b/lightllm/common/triton_utils/autotune_kernel_configs/triton_3.4.0/NVIDIA_H200/grouped_matmul:v1/{K=4096,N=384,expert_num=128,mul_routed_weight=false,out_dtype=torch.bfloat16,topk_num=8,use_fp8_w8a8=false}_NVIDIA_H200.json @@ -0,0 +1,110 @@ +{ + "1": { + "BLOCK_SIZE_K": 128, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 64, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 5, + "num_warps": 4 + }, + "100": { + "BLOCK_SIZE_K": 128, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 32, + "NEED_TRANS": false, + "num_stages": 2, + "num_warps": 4 + }, + "1024": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 128, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 16, + "NEED_TRANS": false, + "num_stages": 5, + "num_warps": 8 + }, + "128": { + "BLOCK_SIZE_K": 128, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 1, + "NEED_TRANS": false, + "num_stages": 2, + "num_warps": 8 + }, + "16": { + "BLOCK_SIZE_K": 128, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 64, + "NEED_TRANS": false, + "num_stages": 4, + "num_warps": 4 + }, + "16384": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 128, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 32, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 8 + }, + "2048": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 64, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 16, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 8 + }, + "256": { + "BLOCK_SIZE_K": 128, + "BLOCK_SIZE_M": 32, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 32, + "NEED_TRANS": false, + "num_stages": 2, + "num_warps": 4 + }, + "32": { + "BLOCK_SIZE_K": 128, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 64, + "GROUP_SIZE_M": 64, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "4096": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 128, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 16, + "NEED_TRANS": false, + "num_stages": 3, + "num_warps": 4 + }, + "64": { + "BLOCK_SIZE_K": 128, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 32, + "NEED_TRANS": false, + "num_stages": 2, + "num_warps": 4 + }, + "8": { + "BLOCK_SIZE_K": 64, + "BLOCK_SIZE_M": 16, + "BLOCK_SIZE_N": 128, + "GROUP_SIZE_M": 32, + "NEED_TRANS": false, + "num_stages": 5, + "num_warps": 4 + } + } \ No newline at end of file