apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

【Fix】 remove text_after_process & raw_prediction (#4421)

Browse Source 
* remove text_after_process &  raw_prediction

* remove text_after_process &  raw_prediction
This commit is contained in:
LiqinruiG
2025-10-16 19:00:18 +08:00
committed by GitHub
parent 8f77adc381
commit 4251ac5e95
20 changed files with 169 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/entrypoints/openai/protocol.py
-8
View File
@@ -193,8 +193,6 @@ class ChatMessage(BaseModel):
tool_calls: Optional[List[DeltaToolCall | ToolCall]] = None
prompt_token_ids: Optional[List[int]] = None
completion_token_ids: Optional[List[int]] = None
text_after_process: Optional[str] = None
raw_prediction: Optional[str] = None
prompt_tokens: Optional[str] = None
completion_tokens: Optional[str] = None
@@ -255,8 +253,6 @@ class DeltaMessage(BaseModel):
completion_token_ids: Optional[List[int]] = None
reasoning_content: Optional[str] = None
tool_calls: Optional[List[DeltaToolCall | ToolCall]] = None
text_after_process: Optional[str] = None
raw_prediction: Optional[str] = None
prompt_tokens: Optional[str] = None
completion_tokens: Optional[str] = None
@@ -295,8 +291,6 @@ class CompletionResponseChoice(BaseModel):
text: str
prompt_token_ids: Optional[List[int]] = None
completion_token_ids: Optional[List[int]] = None
text_after_process: Optional[str] = None
raw_prediction: Optional[str] = None
prompt_tokens: Optional[str] = None
completion_tokens: Optional[str] = None
arrival_time: Optional[float] = None
@@ -341,8 +335,6 @@ class CompletionResponseStreamChoice(BaseModel):
logprobs: Optional[CompletionLogprobs] = None
prompt_token_ids: Optional[List[int]] = None
completion_token_ids: Optional[List[int]] = None
text_after_process: Optional[str] = None
raw_prediction: Optional[str] = None
prompt_tokens: Optional[str] = None
completion_tokens: Optional[str] = None
reasoning_content: Optional[str] = None
fastdeploy/entrypoints/openai/serving_chat.py
+10 -14
View File
@@ -118,14 +118,14 @@ class OpenAIServingChat:
else:
request_id = f"chatcmpl-{uuid.uuid4()}"
api_server_logger.info(f"create chat completion request: {request_id}")
text_after_process = None
prompt_tokens = None
try:
current_req_dict = request.to_dict_for_infer(request_id)
if "chat_template" not in current_req_dict:
current_req_dict["chat_template"] = self.chat_template
current_req_dict["arrival_time"] = time.time()
prompt_token_ids = await self.engine_client.format_and_add_data(current_req_dict)
text_after_process = current_req_dict.get("text_after_process")
prompt_tokens = current_req_dict.get("prompt_tokens")
if isinstance(prompt_token_ids, np.ndarray):
prompt_token_ids = prompt_token_ids.tolist()
except ParameterError as e:
@@ -143,12 +143,12 @@ class OpenAIServingChat:
if request.stream:
return self.chat_completion_stream_generator(
request, request_id, request.model, prompt_token_ids, text_after_process
request, request_id, request.model, prompt_token_ids, prompt_tokens
)
else:
try:
return await self.chat_completion_full_generator(
request, request_id, request.model, prompt_token_ids, text_after_process
request, request_id, request.model, prompt_token_ids, prompt_tokens
)
except Exception as e:
error_msg = f"request[{request_id}]full generator error: {str(e)}, {str(traceback.format_exc())}"
@@ -175,7 +175,7 @@ class OpenAIServingChat:
request_id: str,
model_name: str,
prompt_token_ids: list(),
text_after_process: str,
prompt_tokens: str,
):
"""
Streaming chat completion generator.
@@ -289,8 +289,7 @@ class OpenAIServingChat:
if request.return_token_ids:
choice.delta.prompt_token_ids = list(prompt_token_ids)
choice.delta.text_after_process = text_after_process
choice.delta.prompt_tokens = text_after_process
choice.delta.prompt_tokens = prompt_tokens
chunk = ChatCompletionStreamResponse(
id=request_id,
object=chunk_object_type,
@@ -368,8 +367,7 @@ class OpenAIServingChat:
choice.delta.multimodal_content[0]["completion_token_ids"] = list(output["token_ids"])
else:
choice.delta.completion_token_ids = list(output["token_ids"])
choice.delta.raw_prediction = output.get("raw_prediction")
choice.delta.completion_tokens = output.get("raw_prediction")
choice.delta.completion_tokens = output.get("completion_tokens")
if include_continuous_usage:
chunk.usage = UsageInfo(
prompt_tokens=num_prompt_tokens,
@@ -419,7 +417,7 @@ class OpenAIServingChat:
request_id: str,
model_name: str,
prompt_token_ids: list(),
text_after_process: str,
prompt_tokens: str,
):
"""
Full chat completion generator.
@@ -509,10 +507,8 @@ class OpenAIServingChat:
tool_calls=output.get("tool_call"),
prompt_token_ids=prompt_token_ids if request.return_token_ids else None,
completion_token_ids=completion_token_ids if request.return_token_ids else None,
text_after_process=text_after_process if request.return_token_ids else None,
prompt_tokens=text_after_process if request.return_token_ids else None,
raw_prediction=output.get("raw_prediction") if request.return_token_ids else None,
completion_tokens=output.get("raw_prediction") if request.return_token_ids else None,
prompt_tokens=prompt_tokens if request.return_token_ids else None,
completion_tokens=output.get("completion_tokens") if request.return_token_ids else None,
)
if response_processor.enable_multimodal_content():
fastdeploy/entrypoints/openai/serving_completion.py
+12 -16
View File
@@ -132,7 +132,7 @@ class OpenAIServingCompletion:
num_choices = len(request_prompts)
api_server_logger.info(f"Start preprocessing request: req_id={request_id}), num_choices={num_choices}")
prompt_batched_token_ids = []
text_after_process_list = []
prompt_tokens_list = []
try:
if self.max_waiting_time < 0:
await self.engine_client.semaphore.acquire()
@@ -157,7 +157,7 @@ class OpenAIServingCompletion:
prompt_token_ids = await self.engine_client.format_and_add_data(current_req_dict) # tokenize
if isinstance(prompt_token_ids, np.ndarray):
prompt_token_ids = prompt_token_ids.tolist()
text_after_process_list.append(current_req_dict.get("text_after_process"))
prompt_tokens_list.append(current_req_dict.get("prompt_tokens"))
prompt_batched_token_ids.append(prompt_token_ids)
del current_req_dict
except ParameterError as e:
@@ -180,7 +180,7 @@ class OpenAIServingCompletion:
created_time=created_time,
model_name=request.model,
prompt_batched_token_ids=prompt_batched_token_ids,
text_after_process_list=text_after_process_list,
prompt_tokens_list=prompt_tokens_list,
)
else:
try:
@@ -191,7 +191,7 @@ class OpenAIServingCompletion:
created_time=created_time,
model_name=request.model,
prompt_batched_token_ids=prompt_batched_token_ids,
text_after_process_list=text_after_process_list,
prompt_tokens_list=prompt_tokens_list,
)
except Exception as e:
error_msg = (
@@ -213,7 +213,7 @@ class OpenAIServingCompletion:
created_time: int,
model_name: str,
prompt_batched_token_ids: list(),
text_after_process_list: list(),
prompt_tokens_list: list(),
):
"""
Process the full completion request with multiple choices.
@@ -292,7 +292,7 @@ class OpenAIServingCompletion:
model_name=model_name,
prompt_batched_token_ids=prompt_batched_token_ids,
completion_batched_token_ids=completion_batched_token_ids,
text_after_process_list=text_after_process_list,
prompt_tokens_list=prompt_tokens_list,
)
api_server_logger.info(f"Completion response: {res.model_dump_json()}")
return res
@@ -344,7 +344,7 @@ class OpenAIServingCompletion:
created_time: int,
model_name: str,
prompt_batched_token_ids: list(),
text_after_process_list: list(),
prompt_tokens_list: list(),
):
"""
Process the stream completion request.
@@ -408,8 +408,7 @@ class OpenAIServingCompletion:
index=idx,
text="",
prompt_token_ids=list(prompt_batched_token_ids[idx]),
text_after_process=text_after_process_list[idx],
prompt_tokens=text_after_process_list[idx],
prompt_tokens=prompt_tokens_list[idx],
completion_token_ids=None,
)
],
@@ -443,8 +442,7 @@ class OpenAIServingCompletion:
prompt_token_ids=None,
completion_token_ids=output.get("token_ids") if request.return_token_ids else None,
tool_calls=None,
raw_prediction=output.get("raw_prediction") if request.return_token_ids else None,
completion_tokens=output.get("raw_prediction") if request.return_token_ids else None,
completion_tokens=output.get("completion_tokens") if request.return_token_ids else None,
reasoning_content="",
arrival_time=arrival_time,
logprobs=logprobs_res,
@@ -522,7 +520,7 @@ class OpenAIServingCompletion:
model_name: str,
prompt_batched_token_ids: list(),
completion_batched_token_ids: list(),
text_after_process_list: list(),
prompt_tokens_list: list(),
) -> CompletionResponse:
choices: List[CompletionResponseChoice] = []
num_prompt_tokens = 0
@@ -556,10 +554,8 @@ class OpenAIServingCompletion:
text=output_text,
prompt_token_ids=prompt_token_ids if request.return_token_ids else None,
completion_token_ids=completion_token_ids if request.return_token_ids else None,
raw_prediction=output.get("raw_prediction") if request.return_token_ids else None,
completion_tokens=output.get("raw_prediction") if request.return_token_ids else None,
text_after_process=text_after_process_list[idx] if request.return_token_ids else None,
prompt_tokens=text_after_process_list[idx] if request.return_token_ids else None,
completion_tokens=output.get("completion_tokens") if request.return_token_ids else None,
prompt_tokens=prompt_tokens_list[idx] if request.return_token_ids else None,
reasoning_content=output.get("reasoning_content"),
tool_calls=output.get("tool_call"),
logprobs=aggregated_logprobs,