Fix PD disaggregation warmup: set request_name and improve error logging (#25542)
This commit is contained in:
@@ -1965,6 +1965,7 @@ def _execute_server_warmup(server_args: ServerArgs):
|
|||||||
|
|
||||||
else:
|
else:
|
||||||
logger.info(f"Start of pd disaggregation warmup ...")
|
logger.info(f"Start of pd disaggregation warmup ...")
|
||||||
|
request_name = "/generate"
|
||||||
json_data = {
|
json_data = {
|
||||||
"sampling_params": {
|
"sampling_params": {
|
||||||
"temperature": 0.0,
|
"temperature": 0.0,
|
||||||
@@ -1997,9 +1998,9 @@ def _execute_server_warmup(server_args: ServerArgs):
|
|||||||
_global_state.tokenizer_manager.server_status = ServerStatus.Up
|
_global_state.tokenizer_manager.server_status = ServerStatus.Up
|
||||||
else:
|
else:
|
||||||
logger.info(
|
logger.info(
|
||||||
"Prefill disaggregation mode warm Up Failed, status code: {}".format(
|
"Disaggregation warmup failed (mode=%s), status code: %s",
|
||||||
res.status_code
|
server_args.disaggregation_mode,
|
||||||
)
|
res.status_code,
|
||||||
)
|
)
|
||||||
_global_state.tokenizer_manager.server_status = ServerStatus.UnHealthy
|
_global_state.tokenizer_manager.server_status = ServerStatus.UnHealthy
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user