forked from vllm-project/vllm
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[Kernel][ROCm][AMD] Add fused_moe Triton configs for MI300X (vllm-pro…
…ject#4951) This PR adds Triton kernel configs for the MoE kernel for MI300X
- Loading branch information
1 parent
541ccac
commit c0a626a
Showing
4 changed files
with
494 additions
and
0 deletions.
There are no files selected for viewing
128 changes: 128 additions & 0 deletions
128
.../model_executor/layers/fused_moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI300X.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 +1,128 @@ | ||
| { | ||
| "1": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "2": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "4": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 64, | ||
| "num_stages": 1 | ||
| }, | ||
| "8": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 32, | ||
| "num_stages": 1 | ||
| }, | ||
| "16": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 1 | ||
| }, | ||
| "24": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 64, | ||
| "num_stages": 1 | ||
| }, | ||
| "32": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 1 | ||
| }, | ||
| "48": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 0 | ||
| }, | ||
| "64": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 0 | ||
| }, | ||
| "96": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 16, | ||
| "num_stages": 0 | ||
| }, | ||
| "128": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 0 | ||
| }, | ||
| "256": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 0 | ||
| }, | ||
| "512": { | ||
| "BLOCK_SIZE_M": 256, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 0 | ||
| }, | ||
| "1024": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "1536": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "2048": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "3072": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "4096": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| } | ||
| } |
110 changes: 110 additions & 0 deletions
110
vllm/model_executor/layers/fused_moe/configs/E=8,N=1792,device_name=AMD_Instinct_MI300X.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 +1,110 @@ | ||
| { | ||
| "1": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 64 | ||
| }, | ||
| "2": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 32, | ||
| "GROUP_SIZE_M": 32 | ||
| }, | ||
| "4": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 8 | ||
| }, | ||
| "8": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "16": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "24": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "32": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 8 | ||
| }, | ||
| "48": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 8 | ||
| }, | ||
| "64": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "96": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 8 | ||
| }, | ||
| "128": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 32 | ||
| }, | ||
| "256": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "512": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "1024": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "1536": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "2048": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "3072": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1 | ||
| }, | ||
| "4096": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1 | ||
| } | ||
| } |
128 changes: 128 additions & 0 deletions
128
vllm/model_executor/layers/fused_moe/configs/E=8,N=3584,device_name=AMD_Instinct_MI300X.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 +1,128 @@ | ||
| { | ||
| "1": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 0 | ||
| }, | ||
| "2": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "4": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 32, | ||
| "GROUP_SIZE_M": 32, | ||
| "num_stages": 1 | ||
| }, | ||
| "8": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 1 | ||
| }, | ||
| "16": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 16, | ||
| "num_stages": 1 | ||
| }, | ||
| "24": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 1 | ||
| }, | ||
| "32": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 16, | ||
| "num_stages": 0 | ||
| }, | ||
| "48": { | ||
| "BLOCK_SIZE_M": 16, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 256, | ||
| "GROUP_SIZE_M": 16, | ||
| "num_stages": 1 | ||
| }, | ||
| "64": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 32, | ||
| "num_stages": 0 | ||
| }, | ||
| "96": { | ||
| "BLOCK_SIZE_M": 32, | ||
| "BLOCK_SIZE_N": 32, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 16, | ||
| "num_stages": 0 | ||
| }, | ||
| "128": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 256, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 0 | ||
| }, | ||
| "256": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 8, | ||
| "num_stages": 0 | ||
| }, | ||
| "512": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 64, | ||
| "BLOCK_SIZE_K": 128, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "1024": { | ||
| "BLOCK_SIZE_M": 64, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "1536": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "2048": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "3072": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| }, | ||
| "4096": { | ||
| "BLOCK_SIZE_M": 128, | ||
| "BLOCK_SIZE_N": 128, | ||
| "BLOCK_SIZE_K": 64, | ||
| "GROUP_SIZE_M": 1, | ||
| "num_stages": 0 | ||
| } | ||
| } |
Oops, something went wrong.