chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Fused moe triton cfg opt for rocm (#2612)

Browse Source 
Co-authored-by: wunhuang <wunhuang@amd.com>
This commit is contained in:
kk
2024-12-26 23:38:22 -08:00
committed by GitHub
co-authored by wunhuang
parent c75adfec59
commit 7ca751ff7d
1 changed files with 14 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json
+14 -14
View File
@@ -55,35 +55,35 @@
"kpack": 2
},
"128": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 256,
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_warps": 8,
"num_stages": 0,
"waves_per_eu": 1,
"waves_per_eu": 0,
"matrix_instr_nonkdim": 16,
"kpack": 1
},
"256": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 256,
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_warps": 8,
"num_stages": 0,
"waves_per_eu": 1,
"waves_per_eu": 0,
"matrix_instr_nonkdim": 16,
"kpack": 1
},
"512": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 256,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_warps": 8,
"num_stages": 0,
"waves_per_eu": 2,
"waves_per_eu": 0,
"matrix_instr_nonkdim": 16,
"kpack": 2
},