biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[ROCm][CI] Enable AITER Unified Attention On ROCm For gpt-oss Test (#32431)

Browse Source 
Signed-off-by: Micah Williamson <micah.williamson@amd.com>
This commit is contained in:
Micah Williamson
2026-01-15 18:55:57 -06:00
committed by GitHub
parent bcf2333cd6
commit 46f8a982b1
1 changed files with 15 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
tests/entrypoints/openai/test_serving_chat.py
View File

@@ -10,6 +10,7 @@ import pytest
import pytest_asyncio import pytest_asyncio
from openai import OpenAI from openai import OpenAI
from vllm._aiter_ops import is_aiter_found_and_supported
from vllm.config.multimodal import MultiModalConfig from vllm.config.multimodal import MultiModalConfig
from vllm.entrypoints.openai.chat_completion.protocol import ( from vllm.entrypoints.openai.chat_completion.protocol import (
ChatCompletionRequest, ChatCompletionRequest,
@@ -106,9 +107,21 @@ def gptoss_speculative_server(default_server_args: list[str]):
"--speculative-config", "--speculative-config",
f'{{"model": "{GPT_OSS_SPECULATOR_NAME}", ' f'{{"model": "{GPT_OSS_SPECULATOR_NAME}", '
f'"method": "eagle3", "num_speculative_tokens": 3}}', f'"method": "eagle3", "num_speculative_tokens": 3}}',
"--attention-backend=TRITON_ATTN", f"--attention-backend={
'TRITON_ATTN'
if not is_aiter_found_and_supported()
else 'ROCM_AITER_UNIFIED_ATTN'
}",
] ]
with RemoteOpenAIServer(GPT_OSS_MODEL_NAME, server_args) as remote_server: # gpt-oss requires AITER unified attention on ROCm
# TODO: Remove after fixing TRITON_ATTN issue on ROCm
# https://github.com/vllm-project/vllm/issues/32434
env_dict = None
if is_aiter_found_and_supported():
env_dict = {"VLLM_ROCM_USE_AITER": "1"}
with RemoteOpenAIServer(
GPT_OSS_MODEL_NAME, server_args, env_dict=env_dict
) as remote_server:
yield remote_server yield remote_server