Skip to content

Commit 98d234a

Browse files
bugfix_mla
Signed-off-by: wangxiaoteng <wangxiaoteng@huawei.com>
1 parent dce78b8 commit 98d234a

File tree

1 file changed

+2
-1
lines changed

1 file changed

+2
-1
lines changed

vllm_ascend/worker/worker_v1.py

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -181,7 +181,8 @@ def determine_available_memory(self) -> int:
181181
peak_memory = torch_npu.npu.memory_stats()["allocated_bytes.all.peak"] + \
182182
chuked_prefill_work_space_size
183183
else:
184-
peak_memory = torch_npu.npu.memory_stats()["allocated_bytes.all.peak"]
184+
peak_memory = torch_npu.npu.memory_stats(
185+
)["allocated_bytes.all.peak"]
185186
# TODO: don`t need impl this func after empty_cache in
186187
# Worker.determine_num_available_blocks() unified`
187188
NPUPlatform.empty_cache()

0 commit comments

Comments
 (0)