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

palm streaming

Browse source
This commit is contained in:
ishaan-jaff 2023-09-26 12:27:49 -07:00
parent 5000231eae
commit e7a9246414
2 changed files with 24 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

18
litellm/utils.py
View file

@ -985,6 +985,11 @@ def get_optional_params( # use the openai defaults
optional_params["frequency_penalty"] = frequency_penalty # TODO: Check if should be repetition penalty
if stop != None:
optional_params["stop"] = stop #TG AI expects a list, example ["\n\n\n\n","<|endoftext|>"]
elif custom_llm_provider == "palm":
if temperature != 1:
optional_params["temperature"] = temperature
if top_p != 1:
optional_params["top_p"] = top_p
elif (
model in litellm.vertex_chat_models or model in litellm.vertex_code_chat_models
): # chat-bison has diff args from chat-bison@001, ty Google :)
@ -3088,6 +3093,19 @@ class CustomStreamWrapper:
completion_obj["content"] = new_chunk
self.completion_stream = self.completion_stream[chunk_size:]
time.sleep(0.05)
elif self.custom_llm_provider == "palm":
# fake streaming
if len(self.completion_stream)==0:
if self.sent_last_chunk:
raise StopIteration
else:
model_response.choices[0].finish_reason = "stop"
self.sent_last_chunk = True
chunk_size = 30
new_chunk = self.completion_stream[:chunk_size]
completion_obj["content"] = new_chunk
self.completion_stream = self.completion_stream[chunk_size:]
time.sleep(0.05)
else: # openai chat/azure models
chunk = next(self.completion_stream)
model_response = chunk