[PD] fix: skip warmup request in disaggregation mode to prevent crash on timeout (#5292)
This commit is contained in:
@@ -785,13 +785,17 @@ def _wait_and_warmup(
|
|||||||
json_data["sampling_params"]["max_new_tokens"] = 0
|
json_data["sampling_params"]["max_new_tokens"] = 0
|
||||||
|
|
||||||
try:
|
try:
|
||||||
res = requests.post(
|
if server_args.disaggregation_mode == "null":
|
||||||
url + request_name,
|
res = requests.post(
|
||||||
json=json_data,
|
url + request_name,
|
||||||
headers=headers,
|
json=json_data,
|
||||||
timeout=600,
|
headers=headers,
|
||||||
)
|
timeout=600,
|
||||||
assert res.status_code == 200, f"{res}"
|
)
|
||||||
|
assert res.status_code == 200, f"{res}"
|
||||||
|
else:
|
||||||
|
# Warmup request currently hangs in disaggregation mode, so we skip it.
|
||||||
|
logger.info("Skipping warmup request in disaggregation mode")
|
||||||
except Exception:
|
except Exception:
|
||||||
last_traceback = get_exception_traceback()
|
last_traceback = get_exception_traceback()
|
||||||
if pipe_finish_writer is not None:
|
if pipe_finish_writer is not None:
|
||||||
|
|||||||
Reference in New Issue
Block a user