phoenix/litellm-mirror
0
0
Fork 1
mirror of https://github.com/BerriAI/litellm.git synced 2025-04-26 03:04:13 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

(fix) completion gpt-4 vision check finish_details or finish_reason

Browse source
This commit is contained in:
ishaan-jaff 2023-11-11 10:28:20 -08:00
parent fd6064b571
commit 4d67cee135
1 changed files with 10 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

16
litellm/utils.py
View file

@ -1121,7 +1121,7 @@ def openai_token_counter(messages, model="gpt-3.5-turbo-0613"):
try: try:
encoding = tiktoken.encoding_for_model(model) encoding = tiktoken.encoding_for_model(model)
except KeyError: except KeyError:
print("Warning: model not found. Using cl100k_base encoding.") print_verbose("Warning: model not found. Using cl100k_base encoding.")
encoding = tiktoken.get_encoding("cl100k_base") encoding = tiktoken.get_encoding("cl100k_base")
if model in { if model in {
"gpt-3.5-turbo-0613", "gpt-3.5-turbo-0613",
@ -1137,10 +1137,10 @@ def openai_token_counter(messages, model="gpt-3.5-turbo-0613"):
tokens_per_message = 4 # every message follows <|start|>{role/name}\n{content}<|end|>\n tokens_per_message = 4 # every message follows <|start|>{role/name}\n{content}<|end|>\n
tokens_per_name = -1 # if there's a name, the role is omitted tokens_per_name = -1 # if there's a name, the role is omitted
elif "gpt-3.5-turbo" in model: elif "gpt-3.5-turbo" in model:
print("Warning: gpt-3.5-turbo may update over time. Returning num tokens assuming gpt-3.5-turbo-0613.") print_verbose("Warning: gpt-3.5-turbo may update over time. Returning num tokens assuming gpt-3.5-turbo-0613.")
return openai_token_counter(messages, model="gpt-3.5-turbo-0613") return openai_token_counter(messages, model="gpt-3.5-turbo-0613")
elif "gpt-4" in model: elif "gpt-4" in model:
print("Warning: gpt-4 may update over time. Returning num tokens assuming gpt-4-0613.") print_verbose("Warning: gpt-4 may update over time. Returning num tokens assuming gpt-4-0613.")
return openai_token_counter(messages, model="gpt-4-0613") return openai_token_counter(messages, model="gpt-4-0613")
else: else:
raise NotImplementedError( raise NotImplementedError(
@ -2618,7 +2618,11 @@ def convert_to_model_response_object(response_object: Optional[dict]=None, model
choice_list=[] choice_list=[]
for idx, choice in enumerate(response_object["choices"]): for idx, choice in enumerate(response_object["choices"]):
message = Message(content=choice["message"]["content"], role=choice["message"]["role"], function_call=choice["message"].get("function_call", None)) message = Message(content=choice["message"]["content"], role=choice["message"]["role"], function_call=choice["message"].get("function_call", None))
choice = Choices(finish_reason=choice["finish_reason"], index=idx, message=message) finish_reason = choice.get("finish_reason", None)
if finish_reason == None:
# gpt-4 vision can return 'finish_reason' or 'finish_details'
finish_reason = choice.get("finish_details")
choice = Choices(finish_reason=finish_reason, index=idx, message=message)
choice_list.append(choice) choice_list.append(choice)
model_response_object.choices = choice_list model_response_object.choices = choice_list
@ -2631,8 +2635,8 @@ def convert_to_model_response_object(response_object: Optional[dict]=None, model
if "model" in response_object: if "model" in response_object:
model_response_object.model = response_object["model"] model_response_object.model = response_object["model"]
return model_response_object return model_response_object
except: except Exception as e:
Exception("Invalid response object.") raise Exception(f"Invalid response object {e}")
# NOTE: DEPRECATING this in favor of using success_handler() in Logging: # NOTE: DEPRECATING this in favor of using success_handler() in Logging: