apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-18 22:44:39 +08:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

[Feature] support clear data (#4185)
Some checks failed
CE Compile Job / ce_job_pre_check (push) Has been cancelled
Details
CE Compile Job / print_ce_job_pre_check_outputs (push) Has been cancelled
Details
CE Compile Job / FD-Clone-Linux (push) Has been cancelled
Details
CE Compile Job / Show Code Archive Output (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8090 (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8689 (push) Has been cancelled
Details
CE Compile Job / CE_UPLOAD (push) Has been cancelled
Details

Browse Source 
* fix

* fix

* fix

* [Feature] support clear data

* update

* fix

* fix

* fix

* fix
This commit is contained in:
ltd0924
2025-09-21 20:41:27 +08:00
committed by GitHub
parent 1e86418c4a
commit f75697c2d1
11 changed files with 70 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
fastdeploy/entrypoints/openai/api_server.py
View File

@@ -478,6 +478,7 @@ def reset_scheduler():
if llm_engine is None:
return Response("Engine not loaded", status_code=500)
llm_engine.engine.clear_data()
llm_engine.engine.scheduler.reset()
return Response("Scheduler Reset Successfully", status_code=200)

5
fastdeploy/entrypoints/openai/serving_chat.py
View File

@@ -210,6 +210,8 @@ class OpenAIServingChat:
decoder_base_url=self.tokenizer_base_url,
)
while num_choices > 0:
if self.engine_client.check_model_weight_status():
raise ValueError("Engine is clearing model weight")
try:
response = await asyncio.wait_for(response_queue.get(), timeout=10)
current_waiting_time = 0
@@ -425,6 +427,8 @@ class OpenAIServingChat:
decoder_base_url=self.tokenizer_base_url,
)
while True:
if self.engine_client.check_model_weight_status():
return ErrorResponse(code=400, message="Model weight cleared")
try:
response = await asyncio.wait_for(response_queue.get(), timeout=10)
current_waiting_time = 0
@@ -513,6 +517,7 @@ class OpenAIServingChat:
if final_res.get("error_msg") is not None and "Recover" in final_res["error_msg"]:
choice.finish_reason = "recover_stop"
choices.append(choice)
num_prompt_tokens = len(prompt_token_ids)

5
fastdeploy/entrypoints/openai/serving_completion.py
View File

@@ -216,6 +216,8 @@ class OpenAIServingCompletion:
completion_batched_token_ids = [[] for _ in range(num_choices)]
current_waiting_time = 0
while num_choices > 0:
if self.engine_client.check_model_weight_status():
return ErrorResponse(message="Model weight cleared", code=400)
try:
response = await asyncio.wait_for(response_queue.get(), timeout=10)
current_waiting_time = 0
@@ -270,7 +272,6 @@ class OpenAIServingCompletion:
return res
except Exception as e:
api_server_logger.error(f"Error in completion_full_generator: {e}", exc_info=True)
raise
finally:
self.engine_client.semaphore.release()
if dealer is not None:
@@ -333,6 +334,8 @@ class OpenAIServingCompletion:
)
current_waiting_time = 0
while num_choices > 0:
if self.engine_client.check_model_weight_status():
raise ValueError("Engine is clearing model weight")
try:
response = await asyncio.wait_for(response_queue.get(), timeout=10)
current_waiting_time = 0