phoenix/litellm
1
0
Fork 0
forked from phoenix/litellm-mirror
Code Pull requests Activity

fix(openai.py): drop invalid params if drop_params: true for azure ai

Browse source 
Fixes https://github.com/BerriAI/litellm/issues/4800
This commit is contained in:
Krrish Dholakia 2024-07-20 15:08:15 -07:00
parent 576cccaade
commit 86c9e05c10
2 changed files with 78 additions and 43 deletions
Download patch file Download diff file Expand all files Collapse all files

110
litellm/llms/openai.py
View file

@ -984,47 +984,79 @@ class OpenAIChatCompletion(BaseLLM):
headers=None, headers=None,
): ):
response = None response = None
try: for _ in range(
openai_aclient = self._get_openai_client( 2
is_async=True, ): # if call fails due to alternating messages, retry with reformatted message
api_key=api_key, try:
api_base=api_base, openai_aclient = self._get_openai_client(
timeout=timeout, is_async=True,
max_retries=max_retries, api_key=api_key,
organization=organization, api_base=api_base,
client=client, timeout=timeout,
) max_retries=max_retries,
organization=organization,
client=client,
)
## LOGGING ## LOGGING
logging_obj.pre_call( logging_obj.pre_call(
input=data["messages"], input=data["messages"],
api_key=openai_aclient.api_key, api_key=openai_aclient.api_key,
additional_args={ additional_args={
"headers": {"Authorization": f"Bearer {openai_aclient.api_key}"}, "headers": {
"api_base": openai_aclient._base_url._uri_reference, "Authorization": f"Bearer {openai_aclient.api_key}"
"acompletion": True, },
"complete_input_dict": data, "api_base": openai_aclient._base_url._uri_reference,
}, "acompletion": True,
) "complete_input_dict": data,
},
)
headers, response = await self.make_openai_chat_completion_request( headers, response = await self.make_openai_chat_completion_request(
openai_aclient=openai_aclient, data=data, timeout=timeout openai_aclient=openai_aclient, data=data, timeout=timeout
) )
stringified_response = response.model_dump() stringified_response = response.model_dump()
logging_obj.post_call( logging_obj.post_call(
input=data["messages"], input=data["messages"],
api_key=api_key, api_key=api_key,
original_response=stringified_response, original_response=stringified_response,
additional_args={"complete_input_dict": data}, additional_args={"complete_input_dict": data},
) )
logging_obj.model_call_details["response_headers"] = headers logging_obj.model_call_details["response_headers"] = headers
return convert_to_model_response_object( return convert_to_model_response_object(
response_object=stringified_response, response_object=stringified_response,
model_response_object=model_response, model_response_object=model_response,
hidden_params={"headers": headers}, hidden_params={"headers": headers},
) )
except Exception as e: except openai.UnprocessableEntityError as e:
raise e ## check if body contains unprocessable params - related issue https://github.com/BerriAI/litellm/issues/4800
if litellm.drop_params is True:
if e.body is not None and e.body.get("detail"): # type: ignore
detail = e.body.get("detail") # type: ignore
invalid_params: List[str] = []
if (
isinstance(detail, List)
and len(detail) > 0
and isinstance(detail[0], dict)
):
for error_dict in detail:
if (
error_dict.get("loc")
and isinstance(error_dict.get("loc"), list)
and len(error_dict.get("loc")) == 2
):
invalid_params.append(error_dict["loc"][1])
new_data = {}
for k, v in data.items():
if k not in invalid_params:
new_data[k] = v
data = new_data
else:
raise e
# e.message
except Exception as e:
raise e
def streaming( def streaming(
self, self,

11
litellm/proxy/_new_secret_config.yaml
View file

@ -1,6 +1,9 @@
model_list: model_list:
- model_name: azure-chatgpt - model_name: azure-mistral
litellm_params: litellm_params:
model: azure/chatgpt-v-2 model: azure_ai/mistral
api_key: os.environ/AZURE_API_KEY api_key: os.environ/AZURE_AI_MISTRAL_API_KEY
api_base: os.environ/AZURE_API_BASE api_base: os.environ/AZURE_AI_MISTRAL_API_BASE
litellm_settings:
drop_params: true