phoenix/litellm-mirror
0
0
Fork 1
mirror of https://github.com/BerriAI/litellm.git synced 2025-04-25 18:54:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

(feat) text completion set top_n_tokens for tgi

Browse source
This commit is contained in:
ishaan-jaff 2023-11-01 18:24:04 -07:00
parent 19737f95c5
commit 8ca7af3a63
1 changed files with 14 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

27
litellm/main.py
View file

@ -1821,7 +1821,20 @@ def text_completion(*args, **kwargs):
if "engine" in kwargs: if "engine" in kwargs:
kwargs["model"] = kwargs["engine"] kwargs["model"] = kwargs["engine"]
kwargs.pop("engine") kwargs.pop("engine")
# if echo == True, for TGI llms we need to set top_n_tokens to 3
if kwargs.get("echo", False) == True and (kwargs["model"] not in litellm.open_ai_text_completion_models):
# for tgi llms
if "top_n_tokens" not in kwargs:
kwargs["top_n_tokens"] = 3
if "prompt" in kwargs: if "prompt" in kwargs:
if type(kwargs["prompt"]) == list:
new_prompt = ""
for chunk in kwargs["prompt"]:
decoded = litellm.utils.decode(model="gpt2", tokens=chunk)
new_prompt+= decoded
kwargs["prompt"] = new_prompt
messages = [{"role": "system", "content": kwargs["prompt"]}] messages = [{"role": "system", "content": kwargs["prompt"]}]
kwargs["messages"] = messages kwargs["messages"] = messages
kwargs.pop("prompt") kwargs.pop("prompt")
@ -1834,19 +1847,7 @@ def text_completion(*args, **kwargs):
# only supported for TGI models # only supported for TGI models
try: try:
raw_response = response._hidden_params.get("original_response", None) raw_response = response._hidden_params.get("original_response", None)
tokens = [] transformed_logprobs = litellm.utils.transform_logprobs(raw_response)
token_logprobs = []
if "prefill" in raw_response[0]["details"]:
prefills = raw_response[0]["details"]['prefill']
for prefill in prefills:
tokens.append(prefill['text'])
token_logprobs.append(prefill['logprob'])
new_tokens = [token['text'] for token in raw_response[0]['details']['tokens']]
new_token_logprobs = [token['logprob'] for token in raw_response[0]['details']['tokens']]
transformed_logprobs = {
"tokens": tokens + new_tokens,
"token_logprobs": token_logprobs + new_token_logprobs
}
except Exception as e: except Exception as e:
print("LiteLLM non blocking exception", e) print("LiteLLM non blocking exception", e)
formatted_response_obj = { formatted_response_obj = {