Fix model name included in responses (#24663)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-09-11 10:47:51 -07:00
committed by GitHub
parent 4aa23892d6
commit c1eda615ba
10 changed files with 50 additions and 74 deletions
+1 -1
View File
@@ -186,7 +186,7 @@ class OpenAIServingChat(OpenAIServing):
lora_request = self._maybe_get_adapters(
request, supports_default_mm_loras=True)
model_name = self._get_model_name(request.model, lora_request)
model_name = self.models.model_name(lora_request)
tokenizer = await self.engine_client.get_tokenizer(lora_request)
@@ -146,7 +146,7 @@ class ServingClassification(ClassificationMixin):
request: ClassificationRequest,
raw_request: Request,
) -> Union[ClassificationResponse, ErrorResponse]:
model_name = self._get_model_name(request.model)
model_name = self.models.model_name()
request_id = (f"{self.request_id_prefix}-"
f"{self._base_request_id(raw_request)}")
@@ -232,7 +232,7 @@ class OpenAIServingCompletion(OpenAIServing):
result_generator = merge_async_iterators(*generators)
model_name = self._get_model_name(request.model, lora_request)
model_name = self.models.model_name(lora_request)
num_prompts = len(engine_prompts)
# Similar to the OpenAI API, when n != best_of, we do not stream the
+1 -1
View File
@@ -599,7 +599,7 @@ class OpenAIServingEmbedding(EmbeddingMixin):
See https://platform.openai.com/docs/api-reference/embeddings/create
for the API specification. This API mimics the OpenAI Embedding API.
"""
model_name = self._get_model_name(request.model)
model_name = self.models.model_name()
request_id = (
f"{self.request_id_prefix}-"
f"{self._base_request_id(raw_request, request.request_id)}")
-11
View File
@@ -980,17 +980,6 @@ class OpenAIServing:
return True
return self.models.is_base_model(model_name)
def _get_model_name(
self,
model_name: Optional[str] = None,
lora_request: Optional[LoRARequest] = None,
) -> str:
if lora_request:
return lora_request.lora_name
if not model_name:
return self.models.base_model_paths[0].name
return model_name
def clamp_prompt_logprobs(
prompt_logprobs: Union[PromptLogprobs,
+1 -1
View File
@@ -91,7 +91,7 @@ class OpenAIServingPooling(OpenAIServing):
if error_check_ret is not None:
return error_check_ret
model_name = self._get_model_name(request.model)
model_name = self.models.model_name()
request_id = f"pool-{self._base_request_id(raw_request)}"
created_time = int(time.time())
+1 -1
View File
@@ -237,7 +237,7 @@ class OpenAIServingResponses(OpenAIServing):
try:
lora_request = self._maybe_get_adapters(request)
model_name = self._get_model_name(request.model, lora_request)
model_name = self.models.model_name(lora_request)
tokenizer = await self.engine_client.get_tokenizer(lora_request)
if self.use_harmony:
+2 -2
View File
@@ -353,7 +353,7 @@ class ServingScores(OpenAIServing):
final_res_batch,
request_id,
created_time,
self._get_model_name(request.model),
self.models.model_name(),
)
except asyncio.CancelledError:
return self.create_error_response("Client disconnected")
@@ -399,7 +399,7 @@ class ServingScores(OpenAIServing):
return self.request_output_to_rerank_response(
final_res_batch,
request_id,
self._get_model_name(request.model),
self.models.model_name(),
documents,
top_n,
)