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) textcompletion - transform hf log probs to openai text completion

Browse source
This commit is contained in:
ishaan-jaff 2023-10-31 17:14:37 -07:00
parent b32cedeb28
commit 9223f7cc7a
2 changed files with 11 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

1
litellm/llms/huggingface_restapi.py
View file

@ -348,6 +348,7 @@ def completion(
model_response.usage.completion_tokens = completion_tokens model_response.usage.completion_tokens = completion_tokens
model_response.usage.prompt_tokens = prompt_tokens model_response.usage.prompt_tokens = prompt_tokens
model_response.usage.total_tokens = prompt_tokens + completion_tokens model_response.usage.total_tokens = prompt_tokens + completion_tokens
model_response._hidden_params["original_response"] = completion_response
return model_response return model_response

11
litellm/main.py
View file

@ -1791,6 +1791,15 @@ def text_completion(*args, **kwargs):
# if the model is text-davinci-003, return raw response from openai # if the model is text-davinci-003, return raw response from openai
if kwargs["model"] in litellm.open_ai_text_completion_models and response._hidden_params.get("original_response", None) != None: if kwargs["model"] in litellm.open_ai_text_completion_models and response._hidden_params.get("original_response", None) != None:
return response._hidden_params.get("original_response", None) return response._hidden_params.get("original_response", None)
transformed_logprobs = None
try:
raw_response = response._hidden_params.get("original_response", None)
transformed_logprobs = {
"tokens": [token['text'] for token in raw_response[0]['details']['tokens']],
"token_logprobs": [token['logprob'] for token in raw_response[0]['details']['tokens']]
}
except Exception as e:
print("LiteLLM non blocking exception", e)
formatted_response_obj = { formatted_response_obj = {
"id": response["id"], "id": response["id"],
"object": "text_completion", "object": "text_completion",
@ -1800,7 +1809,7 @@ def text_completion(*args, **kwargs):
{ {
"text": response["choices"][0]["message"]["content"], "text": response["choices"][0]["message"]["content"],
"index": response["choices"][0]["index"], "index": response["choices"][0]["index"],
"logprobs": None, "logprobs": transformed_logprobs,
"finish_reason": response["choices"][0]["finish_reason"] "finish_reason": response["choices"][0]["finish_reason"]
} }
], ],