EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[Bug Fix] Fix bug of platform for parameter checking (#411)

Browse Source 
Fix bug in platform.py to avoid the None value of config parameters.

Signed-off-by: wuhuikx <wuhui_csu@163.com>
This commit is contained in:
wuhuikx
2025-03-28 16:31:27 +08:00
committed by GitHub
parent b1557abab6
commit 57a84bb7be
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm_ascend/platform.py
View File

@@ -89,14 +89,14 @@ class NPUPlatform(Platform):
@classmethod @classmethod
def check_and_update_config(cls, vllm_config: VllmConfig) -> None: def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
compilation_config = vllm_config.compilation_config compilation_config = vllm_config.compilation_config
if compilation_config.level != CompilationLevel.NO_COMPILATION: if compilation_config and compilation_config.level != CompilationLevel.NO_COMPILATION:
logger.warning( logger.warning(
"Compilation level %s is not supported on NPU now, forcing compilation level to NO_COMPILATION", "Compilation level %s is not supported on NPU now, forcing compilation level to NO_COMPILATION",
compilation_config.level) compilation_config.level)
compilation_config.level = CompilationLevel.NO_COMPILATION compilation_config.level = CompilationLevel.NO_COMPILATION
parallel_config = vllm_config.parallel_config parallel_config = vllm_config.parallel_config
if parallel_config.worker_cls == "auto": if parallel_config and parallel_config.worker_cls == "auto":
if envs.VLLM_USE_V1: if envs.VLLM_USE_V1:
parallel_config.worker_cls = "vllm_ascend.worker.worker_v1.NPUWorker" parallel_config.worker_cls = "vllm_ascend.worker.worker_v1.NPUWorker"
elif vllm_config.speculative_config: elif vllm_config.speculative_config:
@@ -111,7 +111,7 @@ class NPUPlatform(Platform):
if cache_config and cache_config.block_size is None: if cache_config and cache_config.block_size is None:
cache_config.block_size = 128 cache_config.block_size = 128
if envs.VLLM_USE_V1 and cache_config.enable_prefix_caching: if envs.VLLM_USE_V1 and cache_config and cache_config.enable_prefix_caching:
logger.warning( logger.warning(
"Prefix caching is not supported for V1 now, disable prefix caching" "Prefix caching is not supported for V1 now, disable prefix caching"
) )