mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2025-12-13 23:45:01 +08:00
[FEAT] [ROCm]: Add Qwen/Qwen3-235B-A22B-FP8 TP4 triton fused moe config (#17535)
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
This commit is contained in:
parent
7169f87ad0
commit
f5a3c655b2
@ -0,0 +1,164 @@
|
|||||||
|
{
|
||||||
|
"1": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"2": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"4": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"8": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"16": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"24": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"32": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"48": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"64": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 4,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"96": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"128": {
|
||||||
|
"BLOCK_SIZE_M": 16,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"256": {
|
||||||
|
"BLOCK_SIZE_M": 32,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 4,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"512": {
|
||||||
|
"BLOCK_SIZE_M": 64,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 8,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"1024": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 8,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"1536": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 8,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"2048": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 128,
|
||||||
|
"GROUP_SIZE_M": 4,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"3072": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 1,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
},
|
||||||
|
"4096": {
|
||||||
|
"BLOCK_SIZE_M": 128,
|
||||||
|
"BLOCK_SIZE_N": 128,
|
||||||
|
"BLOCK_SIZE_K": 256,
|
||||||
|
"GROUP_SIZE_M": 4,
|
||||||
|
"num_warps": 8,
|
||||||
|
"num_stages": 2,
|
||||||
|
"waves_per_eu": 0
|
||||||
|
}
|
||||||
|
}
|
||||||
Loading…
x
Reference in New Issue
Block a user