From fefdc38b841f54863e34d662569dcf0c4f022343 Mon Sep 17 00:00:00 2001 From: Jade Zheng Date: Fri, 13 Jun 2025 01:52:47 +0800 Subject: [PATCH 1/5] fix: add validation for expert parallelism settings Add validation to prevent simultaneous use of `--enable-expert-parallel` and `expert-tensor-parallel-size` configurations. These settings are mutually exclusive. Implementing this check prevents unexpected behavior and improves error tracing. If both settings are enabled concurrently, the system now throws an error, making it easier to identify and resolve configuration issues. Signed-off-by: Jade Zheng --- vllm_ascend/ascend_config.py | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) diff --git a/vllm_ascend/ascend_config.py b/vllm_ascend/ascend_config.py index 2d3428327..a3c45c91d 100644 --- a/vllm_ascend/ascend_config.py +++ b/vllm_ascend/ascend_config.py @@ -13,11 +13,14 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -from typing import Optional +from typing import Optional, TYPE_CHECKING import vllm.envs as envs from vllm.logger import logger +if TYPE_CHECKING: + from vllm.config import VllmConfig + class AscendConfig: """ @@ -112,7 +115,7 @@ def get_ascend_config(): return _ASCEND_CONFIG -def check_ascend_config(vllm_config, enforce_eager): +def check_ascend_config(vllm_config: "VllmConfig", enforce_eager): ascend_config = get_ascend_config() # for v0 engine @@ -164,3 +167,11 @@ def check_ascend_config(vllm_config, enforce_eager): "ACL Graph is currently experimental. Please " "raise an issue on https://github.com/vllm-project/vllm-ascend/issues" " if you encourage any Error") + + # for expert parallelism + if vllm_config.parallel_config.enable_expert_parallel and \ + ascend_config.expert_tensor_parallel_size > 0: + raise RuntimeError( + "Cannot set `--enable-expert-parallel` and " + "`expert_tensor_parallel_size` at the same time. " + ) From de9230c92e168b401413626c26a898b7866ef447 Mon Sep 17 00:00:00 2001 From: Jade Zheng Date: Fri, 13 Jun 2025 01:53:47 +0800 Subject: [PATCH 2/5] chore: lint Signed-off-by: Jade Zheng --- vllm_ascend/ascend_config.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm_ascend/ascend_config.py b/vllm_ascend/ascend_config.py index a3c45c91d..5d2f9096e 100644 --- a/vllm_ascend/ascend_config.py +++ b/vllm_ascend/ascend_config.py @@ -13,7 +13,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -from typing import Optional, TYPE_CHECKING +from typing import TYPE_CHECKING, Optional import vllm.envs as envs from vllm.logger import logger From bc3d01c846071d1f19faf4c6f9834fe3d81e3059 Mon Sep 17 00:00:00 2001 From: Jade Zheng Date: Fri, 13 Jun 2025 01:55:59 +0800 Subject: [PATCH 3/5] update Signed-off-by: Jade Zheng --- vllm_ascend/ascend_config.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/vllm_ascend/ascend_config.py b/vllm_ascend/ascend_config.py index 5d2f9096e..6bcf9cfc0 100644 --- a/vllm_ascend/ascend_config.py +++ b/vllm_ascend/ascend_config.py @@ -170,8 +170,8 @@ def check_ascend_config(vllm_config: "VllmConfig", enforce_eager): # for expert parallelism if vllm_config.parallel_config.enable_expert_parallel and \ - ascend_config.expert_tensor_parallel_size > 0: + ascend_config.expert_tensor_parallel_size > 1: raise RuntimeError( "Cannot set `--enable-expert-parallel` and " - "`expert_tensor_parallel_size` at the same time. " + "`expert_tensor_parallel_size` > 1 at the same time." ) From 6322681810ca9efc31a25d2b49e8986ceaddfc02 Mon Sep 17 00:00:00 2001 From: Jade Zheng Date: Fri, 13 Jun 2025 10:19:48 +0800 Subject: [PATCH 4/5] chore: lint Signed-off-by: Jade Zheng --- vllm_ascend/ascend_config.py | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/vllm_ascend/ascend_config.py b/vllm_ascend/ascend_config.py index 6bcf9cfc0..9c2ba0462 100644 --- a/vllm_ascend/ascend_config.py +++ b/vllm_ascend/ascend_config.py @@ -173,5 +173,4 @@ def check_ascend_config(vllm_config: "VllmConfig", enforce_eager): ascend_config.expert_tensor_parallel_size > 1: raise RuntimeError( "Cannot set `--enable-expert-parallel` and " - "`expert_tensor_parallel_size` > 1 at the same time." - ) + "`expert_tensor_parallel_size` > 1 at the same time.") From f66efbbc1233c28a89b939da5007824119622885 Mon Sep 17 00:00:00 2001 From: Jade Zheng Date: Sun, 15 Jun 2025 15:38:43 +0800 Subject: [PATCH 5/5] Update vllm_ascend/ascend_config.py Signed-off-by: Jade Zheng Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> --- vllm_ascend/ascend_config.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm_ascend/ascend_config.py b/vllm_ascend/ascend_config.py index 9c2ba0462..500f38eee 100644 --- a/vllm_ascend/ascend_config.py +++ b/vllm_ascend/ascend_config.py @@ -171,6 +171,6 @@ def check_ascend_config(vllm_config: "VllmConfig", enforce_eager): # for expert parallelism if vllm_config.parallel_config.enable_expert_parallel and \ ascend_config.expert_tensor_parallel_size > 1: - raise RuntimeError( + raise ValueError( "Cannot set `--enable-expert-parallel` and " "`expert_tensor_parallel_size` > 1 at the same time.")