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

[bugfix][0.11]fix proxy decode bug (#3751)

Browse Source 
### What this PR does / why we need it?
fix proxy decode bug while parsing non-UTF-8 characters.

---------

Signed-off-by: CHEN <116010019@link.cuhk.edu.cn>
This commit is contained in:
Shirley125
2025-10-27 16:56:50 +08:00
committed by GitHub
parent 43276fd822
commit e48ca0b6ec
2 changed files with 14 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
examples/disaggregated_prefill_v1/load_balance_proxy_layerwise_server_example.py
View File

@@ -556,7 +556,13 @@ async def _handle_completions(api: str, request: Request):
instance_info.prefiller_idx, instance_info.prefiller_idx,
instance_info.prefiller_score) instance_info.prefiller_score)
released_kv = True released_kv = True
chunk_str = chunk.decode("utf-8").strip() try:
chunk_str = chunk.decode("utf-8").strip()
except UnicodeDecodeError:
logger.debug(
f"Skipping chunk: {chunk}")
yield chunk
continue
if not chunk_str: if not chunk_str:
continue continue
if chunk_str.startswith("data: "): if chunk_str.startswith("data: "):

8
examples/disaggregated_prefill_v1/load_balance_proxy_server_example.py
View File

@@ -539,7 +539,13 @@ async def _handle_completions(api: str, request: Request):
instance_info.prefiller_idx, instance_info.prefiller_idx,
instance_info.prefiller_score) instance_info.prefiller_score)
released_kv = True released_kv = True
chunk_str = chunk.decode("utf-8").strip() try:
chunk_str = chunk.decode("utf-8").strip()
except UnicodeDecodeError:
logger.debug(
f"Skipping chunk: {chunk}")
yield chunk
continue
if not chunk_str: if not chunk_str:
continue continue
if chunk_str.startswith("data: "): if chunk_str.startswith("data: "):