Fused moe triton cfg opt for rocm (#2612)

Co-authored-by: wunhuang <wunhuang@amd.com>
This commit is contained in:
kk
2024-12-27 15:38:22 +08:00
committed by GitHub
parent c75adfec59
commit 7ca751ff7d

View File

@@ -55,35 +55,35 @@
"kpack": 2 "kpack": 2
}, },
"128": { "128": {
"BLOCK_SIZE_M": 16, "BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 64, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 256, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1, "GROUP_SIZE_M": 1,
"num_warps": 4, "num_warps": 8,
"num_stages": 0, "num_stages": 0,
"waves_per_eu": 1, "waves_per_eu": 0,
"matrix_instr_nonkdim": 16, "matrix_instr_nonkdim": 16,
"kpack": 1 "kpack": 1
}, },
"256": { "256": {
"BLOCK_SIZE_M": 16, "BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 64, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 256, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1, "GROUP_SIZE_M": 1,
"num_warps": 4, "num_warps": 8,
"num_stages": 0, "num_stages": 0,
"waves_per_eu": 1, "waves_per_eu": 0,
"matrix_instr_nonkdim": 16, "matrix_instr_nonkdim": 16,
"kpack": 1 "kpack": 1
}, },
"512": { "512": {
"BLOCK_SIZE_M": 64, "BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 64, "BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 256, "BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1, "GROUP_SIZE_M": 1,
"num_warps": 4, "num_warps": 8,
"num_stages": 0, "num_stages": 0,
"waves_per_eu": 2, "waves_per_eu": 0,
"matrix_instr_nonkdim": 16, "matrix_instr_nonkdim": 16,
"kpack": 2 "kpack": 2
}, },