mirror of
https://github.com/BerriAI/litellm.git
synced 2025-04-25 10:44:24 +00:00
openai at parity
This commit is contained in:
ishaan-jaff
parent
1739cbdd6c
commit
cfddfaff89
2 changed files with 209 additions and 32 deletions
124
litellm/main.py
124
litellm/main.py
|
|
@ -82,12 +82,69 @@ def client(original_function):
|
|||
raise e
|
||||
return wrapper
|
||||
|
||||
|
||||
def get_optional_params(
|
||||
# 12 optional params
|
||||
functions = [],
|
||||
function_call = "",
|
||||
temperature = 1,
|
||||
top_p = 1,
|
||||
n = 1,
|
||||
stream = False,
|
||||
stop = None,
|
||||
max_tokens = float('inf'),
|
||||
presence_penalty = 0,
|
||||
frequency_penalty = 0,
|
||||
logit_bias = {},
|
||||
user = "",
|
||||
):
|
||||
optional_params = {}
|
||||
if functions != []:
|
||||
optional_params["functions"] = functions
|
||||
if function_call != "":
|
||||
optional_params["function_call"] = function_call
|
||||
if temperature != 1:
|
||||
optional_params["temperature"] = temperature
|
||||
if top_p != 1:
|
||||
optional_params["top_p"] = top_p
|
||||
if n != 1:
|
||||
optional_params["n"] = n
|
||||
if stream:
|
||||
optional_params["stream"] = stream
|
||||
if stop != None:
|
||||
optional_params["stop"] = stop
|
||||
if max_tokens != float('inf'):
|
||||
optional_params["max_tokens"] = max_tokens
|
||||
if presence_penalty != 0:
|
||||
optional_params["presence_penalty"] = presence_penalty
|
||||
if frequency_penalty != 0:
|
||||
optional_params["frequency_penalty"] = frequency_penalty
|
||||
if logit_bias != {}:
|
||||
optional_params["logit_bias"] = logit_bias
|
||||
if user != "":
|
||||
optional_params["user"] = user
|
||||
return optional_params
|
||||
|
||||
####### COMPLETION ENDPOINTS ################
|
||||
#############################################
|
||||
@client
|
||||
@func_set_timeout(60, allowOverride=True) ## https://pypi.org/project/func-timeout/ - timeouts, in case calls hang (e.g. Azure)
|
||||
def completion(model, messages, max_tokens=None, *, forceTimeout=60, azure=False, logger_fn=None): # ,*,.. requires optional params like forceTimeout, azure and logger_fn to be passed in as keyword arguments
|
||||
def completion(
|
||||
model, messages, # required params
|
||||
# Optional OpenAI params: see https://platform.openai.com/docs/api-reference/chat/create
|
||||
functions=[], function_call="", # optional params
|
||||
temperature=1, top_p=1, n=1, stream=False, stop=None, max_tokens=float('inf'),
|
||||
presence_penalty=0, frequency_penalty=0, logit_bias={}, user="",
|
||||
# Optional liteLLM function params
|
||||
azure=False, logger_fn=None, verbose=False
|
||||
):
|
||||
try:
|
||||
# check if user passed in any of the OpenAI optional params
|
||||
optional_params = get_optional_params(
|
||||
functions=functions, function_call=function_call,
|
||||
temperature=temperature, top_p=top_p, n=n, stream=stream, stop=stop, max_tokens=max_tokens,
|
||||
presence_penalty=presence_penalty, frequency_penalty=frequency_penalty, logit_bias=logit_bias, user=user
|
||||
)
|
||||
if azure == True:
|
||||
# azure configs
|
||||
openai.api_type = "azure"
|
||||
|
|
@ -95,11 +152,39 @@ def completion(model, messages, max_tokens=None, *, forceTimeout=60, azure=False
|
|||
openai.api_version = os.environ.get("AZURE_API_VERSION")
|
||||
openai.api_key = os.environ.get("AZURE_API_KEY")
|
||||
## LOGGING
|
||||
logging(model=model, input=input, azure=azure, logger_fn=logger_fn)
|
||||
logging(model=model, input=messages, azure=azure, logger_fn=logger_fn)
|
||||
## COMPLETION CALL
|
||||
response = openai.ChatCompletion.create(
|
||||
engine=model,
|
||||
messages = messages
|
||||
messages = messages,
|
||||
**optional_params
|
||||
)
|
||||
elif model in open_ai_chat_completion_models:
|
||||
openai.api_type = "openai"
|
||||
openai.api_base = "https://api.openai.com/v1"
|
||||
openai.api_version = None
|
||||
openai.api_key = os.environ.get("OPENAI_API_KEY")
|
||||
## LOGGING
|
||||
logging(model=model, input=messages, azure=azure, logger_fn=logger_fn)
|
||||
|
||||
## COMPLETION CALL
|
||||
response = openai.ChatCompletion.create(
|
||||
model=model,
|
||||
messages = messages,
|
||||
**optional_params
|
||||
)
|
||||
elif model in open_ai_text_completion_models:
|
||||
openai.api_type = "openai"
|
||||
openai.api_base = "https://api.openai.com/v1"
|
||||
openai.api_version = None
|
||||
openai.api_key = os.environ.get("OPENAI_API_KEY")
|
||||
prompt = " ".join([message["content"] for message in messages])
|
||||
## LOGGING
|
||||
logging(model=model, input=prompt, azure=azure, logger_fn=logger_fn)
|
||||
## COMPLETION CALL
|
||||
response = openai.Completion.create(
|
||||
model=model,
|
||||
prompt = prompt
|
||||
)
|
||||
elif "replicate" in model:
|
||||
# replicate defaults to os.environ.get("REPLICATE_API_TOKEN")
|
||||
|
|
@ -109,7 +194,7 @@ def completion(model, messages, max_tokens=None, *, forceTimeout=60, azure=False
|
|||
os.environ["REPLICATE_API_TOKEN"] = replicate_api_token
|
||||
prompt = " ".join([message["content"] for message in messages])
|
||||
input = {"prompt": prompt}
|
||||
if max_tokens:
|
||||
if max_tokens != float('inf'):
|
||||
input["max_length"] = max_tokens # for t5 models
|
||||
input["max_new_tokens"] = max_tokens # for llama2 models
|
||||
## LOGGING
|
||||
|
|
@ -147,9 +232,10 @@ def completion(model, messages, max_tokens=None, *, forceTimeout=60, azure=False
|
|||
prompt += f"{HUMAN_PROMPT}{message['content']}"
|
||||
prompt += f"{AI_PROMPT}"
|
||||
anthropic = Anthropic()
|
||||
if max_tokens:
|
||||
# check if user passed in max_tokens != float('inf')
|
||||
if max_tokens != float('inf'):
|
||||
max_tokens_to_sample = max_tokens
|
||||
else:
|
||||
else:
|
||||
max_tokens_to_sample = 300 # default in Anthropic docs https://docs.anthropic.com/claude/reference/client-libraries
|
||||
## LOGGING
|
||||
logging(model=model, input=prompt, azure=azure, additional_args={"max_tokens": max_tokens}, logger_fn=logger_fn)
|
||||
|
|
@ -197,32 +283,6 @@ def completion(model, messages, max_tokens=None, *, forceTimeout=60, azure=False
|
|||
],
|
||||
}
|
||||
response = new_response
|
||||
|
||||
elif model in open_ai_chat_completion_models:
|
||||
openai.api_type = "openai"
|
||||
openai.api_base = "https://api.openai.com/v1"
|
||||
openai.api_version = None
|
||||
openai.api_key = os.environ.get("OPENAI_API_KEY")
|
||||
## LOGGING
|
||||
logging(model=model, input=messages, azure=azure, logger_fn=logger_fn)
|
||||
## COMPLETION CALL
|
||||
response = openai.ChatCompletion.create(
|
||||
model=model,
|
||||
messages = messages
|
||||
)
|
||||
elif model in open_ai_text_completion_models:
|
||||
openai.api_type = "openai"
|
||||
openai.api_base = "https://api.openai.com/v1"
|
||||
openai.api_version = None
|
||||
openai.api_key = os.environ.get("OPENAI_API_KEY")
|
||||
prompt = " ".join([message["content"] for message in messages])
|
||||
## LOGGING
|
||||
logging(model=model, input=prompt, azure=azure, logger_fn=logger_fn)
|
||||
## COMPLETION CALL
|
||||
response = openai.Completion.create(
|
||||
model=model,
|
||||
prompt = prompt
|
||||
)
|
||||
else:
|
||||
logging(model=model, input=messages, azure=azure, logger_fn=logger_fn)
|
||||
return response
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue