Skip to content

format code; add deepseekv3 tuning config #6

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tjtanaa merged 1 commit into from
Mar 8, 2025
Merged

format code; add deepseekv3 tuning config #6

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions benchmarks/kernels/benchmark_mixtral_moe_rocm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
import argparse
import json
import os
Expand Down
1 change: 1 addition & 0 deletions benchmarks/test_accuracy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
import argparse
import dataclasses

Expand Down
1 change: 1 addition & 0 deletions tests/kernels/test_semi_structured.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
"""Tests for sparse cutlass kernels

Run `pytest tests/kernels/test_semi_structured.py`.
Expand Down
1 change: 1 addition & 0 deletions vllm/attention/ops/paged_attn_ater.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
from dataclasses import dataclass
from typing import List, Optional, Tuple

Expand Down
5 changes: 3 additions & 2 deletions vllm/entrypoints/fast_sync_llm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
import multiprocessing as mp
from queue import Empty
from typing import Union
Expand Down Expand Up @@ -85,8 +86,8 @@ def run_engine(self):
try:
while True:
poll_interval -= 1
if (self.input_queue.qsize() >=
envs.VLLM_SYNC_SERVER_ACCUM_REQUESTS
if (self.input_queue.qsize()
>= envs.VLLM_SYNC_SERVER_ACCUM_REQUESTS
or poll_interval <= 0
or not self.llm_engine.has_unfinished_requests()):
self._poll_requests()
Expand Down
1 change: 1 addition & 0 deletions vllm/entrypoints/sync_openai/api_server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
import asyncio
import multiprocessing
import re
Expand Down
164 changes: 164 additions & 0 deletions ...used_moe/configs/E=256,N=256,device_name=MI308X,dtype=fp8_w8a8,block_shape=[128,128].json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,164 @@
{
"1": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 256,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"2": {
"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
},
"4": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 2,
"num_stages": 2,
"waves_per_eu": 0
},
"8": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 2,
"num_stages": 2,
"waves_per_eu": 0
},
"16": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"24": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"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": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"48": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"64": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"96": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"128": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"256": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"512": {
"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
},
"1024": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 4,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"1536": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 4,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"2048": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
},
"3072": {
"BLOCK_SIZE_M": 128,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 4,
"num_warps": 8,
"num_stages": 2,
"waves_per_eu": 0
},
"4096": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 256,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0
}
}
41 changes: 15 additions & 26 deletions vllm/model_executor/layers/fused_moe/configs/E=8,N=8192,device_name=MI308X.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"1": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 2,
Expand All @@ -19,11 +19,11 @@
"num_stages": 2,
"waves_per_eu": 0,
"matrix_instr_nonkdim": 16,
"kpack": 2
"kpack": 1
},
"4": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 2,
Expand All @@ -34,7 +34,7 @@
},
"8": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 2,
Expand All @@ -45,7 +45,7 @@
},
"16": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 2,
Expand All @@ -56,7 +56,7 @@
},
"24": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 2,
Expand All @@ -78,7 +78,7 @@
},
"48": {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 2,
Expand All @@ -89,8 +89,8 @@
},
"64": {
"BLOCK_SIZE_M": 32,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 256,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 4,
"num_warps": 4,
"num_stages": 2,
Expand All @@ -100,8 +100,8 @@
},
"96": {
"BLOCK_SIZE_M": 32,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 256,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 4,
"num_warps": 4,
"num_stages": 2,
Expand All @@ -123,19 +123,8 @@
"256": {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 4,
"num_warps": 4,
"num_stages": 2,
"waves_per_eu": 0,
"matrix_instr_nonkdim": 32,
"kpack": 2
},
"256": {
"BLOCK_SIZE_M": 128,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 4,
"BLOCK_SIZE_K": 128,
"GROUP_SIZE_M": 1,
"num_warps": 8,
"num_stages": 2,
"waves_per_eu": 0,
Expand All @@ -151,13 +140,13 @@
"num_stages": 2,
"waves_per_eu": 0,
"matrix_instr_nonkdim": 16,
"kpack": 2
"kpack": 1
},
"1024": {
"BLOCK_SIZE_M": 128,
"BLOCK_SIZE_N": 128,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"GROUP_SIZE_M": 4,
"num_warps": 8,
"num_stages": 2,
"waves_per_eu": 0,
Expand Down
Loading