-
-
Notifications
You must be signed in to change notification settings - Fork 4.8k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[Kernel] adding fused moe kernel config for L40S TP4 (#9245)
- Loading branch information
Showing
1 changed file
with
173 additions
and
0 deletions.
There are no files selected for viewing
173 changes: 173 additions & 0 deletions
173
vllm/model_executor/layers/fused_moe/configs/E=8,N=3584,device_name=NVIDIA_L40S.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,173 @@ | ||
| { | ||
| "1": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_warps": 4, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "2": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 32, | ||
| "GROUP_SIZE_M": 2, | ||
| "num_warps": 4, | ||
| "num_ctas": 1, | ||
| "num_stages": 7 | ||
| }, | ||
| "4": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 32, | ||
| "GROUP_SIZE_M": 128, | ||
| "num_warps": 2, | ||
| "num_ctas": 1, | ||
| "num_stages": 4 | ||
| }, | ||
| "8": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_warps": 2, | ||
| "num_ctas": 1, | ||
| "num_stages": 1 | ||
| }, | ||
| "16": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_warps": 2, | ||
| "num_ctas": 1, | ||
| "num_stages": 1 | ||
| }, | ||
| "32": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 2, | ||
| "num_warps": 4, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "48": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 2, | ||
| "num_warps": 4, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "64": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 2, | ||
| "num_warps": 4, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "96": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_warps": 8, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "128": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_warps": 8, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "192": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_warps": 8, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "256": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_warps": 16, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "512": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 128, | ||
| "num_warps": 2, | ||
| "num_ctas": 1, | ||
| "num_stages": 8 | ||
| }, | ||
| "1024": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 4, | ||
| "num_warps": 8, | ||
| "num_ctas": 1, | ||
| "num_stages": 3 | ||
| }, | ||
| "2048": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_warps": 16, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "3072": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_warps": 16, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "4096": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 32, | ||
| "GROUP_SIZE_M": 64, | ||
| "num_warps": 8, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "6144": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 32, | ||
| "GROUP_SIZE_M": 64, | ||
| "num_warps": 8, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| }, | ||
| "8192": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_warps": 16, | ||
| "num_ctas": 1, | ||
| "num_stages": 2 | ||
| } | ||
| } |