EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Fix openai protocols and pass top_k, min_p (#2499)

Browse Source
This commit is contained in:
Lianmin Zheng
2024-12-17 04:14:14 -08:00
committed by GitHub
parent 33c5ff2845
commit 361ea8d912
2 changed files with 13 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
python/sglang/srt/openai_api/adapter.py
View File

@@ -510,6 +510,8 @@ def v1_generate_request(
"stop": request.stop, "stop": request.stop,
"stop_token_ids": request.stop_token_ids, "stop_token_ids": request.stop_token_ids,
"top_p": request.top_p, "top_p": request.top_p,
"top_k": request.top_k,
"min_p": request.min_p,
"presence_penalty": request.presence_penalty, "presence_penalty": request.presence_penalty,
"frequency_penalty": request.frequency_penalty, "frequency_penalty": request.frequency_penalty,
"repetition_penalty": request.repetition_penalty, "repetition_penalty": request.repetition_penalty,
@@ -926,6 +928,8 @@ def v1_chat_generate_request(
"stop": stop, "stop": stop,
"stop_token_ids": request.stop_token_ids, "stop_token_ids": request.stop_token_ids,
"top_p": request.top_p, "top_p": request.top_p,
"top_k": request.top_k,
"min_p": request.min_p,
"presence_penalty": request.presence_penalty, "presence_penalty": request.presence_penalty,
"frequency_penalty": request.frequency_penalty, "frequency_penalty": request.frequency_penalty,
"repetition_penalty": request.repetition_penalty, "repetition_penalty": request.repetition_penalty,

14
python/sglang/srt/openai_api/protocol.py
View File

@@ -166,17 +166,19 @@ class CompletionRequest(BaseModel):
temperature: float = 1.0 temperature: float = 1.0
top_p: float = 1.0 top_p: float = 1.0
user: Optional[str] = None user: Optional[str] = None
lora_path: Optional[Union[List[Optional[str]], Optional[str]]] = None
# Extra parameters for SRT backend only and will be ignored by OpenAI models. # Extra parameters for SRT backend only and will be ignored by OpenAI models.
json_schema: Optional[str] = None top_k: int = -1
regex: Optional[str] = None min_p: float = 0.0
min_tokens: int = 0 min_tokens: int = 0
regex: Optional[str] = None
json_schema: Optional[str] = None
repetition_penalty: float = 1.0 repetition_penalty: float = 1.0
stop_token_ids: Optional[List[int]] = None stop_token_ids: Optional[List[int]] = None
no_stop_trim: bool = False no_stop_trim: bool = False
ignore_eos: bool = False ignore_eos: bool = False
skip_special_tokens: bool = True skip_special_tokens: bool = True
lora_path: Optional[Union[List[Optional[str]], Optional[str]]] = None
class CompletionResponseChoice(BaseModel): class CompletionResponseChoice(BaseModel):
@@ -276,14 +278,16 @@ class ChatCompletionRequest(BaseModel):
user: Optional[str] = None user: Optional[str] = None
# Extra parameters for SRT backend only and will be ignored by OpenAI models. # Extra parameters for SRT backend only and will be ignored by OpenAI models.
lora_path: Optional[Union[List[Optional[str]], Optional[str]]] = None top_k: int = -1
regex: Optional[str] = None min_p: float = 0.0
min_tokens: int = 0 min_tokens: int = 0
regex: Optional[str] = None
repetition_penalty: float = 1.0 repetition_penalty: float = 1.0
stop_token_ids: Optional[List[int]] = None stop_token_ids: Optional[List[int]] = None
no_stop_trim: bool = False no_stop_trim: bool = False
ignore_eos: bool = False ignore_eos: bool = False
skip_special_tokens: bool = True skip_special_tokens: bool = True
lora_path: Optional[Union[List[Optional[str]], Optional[str]]] = None
class ChatMessage(BaseModel): class ChatMessage(BaseModel):