Output log probabilities

Check your docs version

These docs are for the new Anyscale design. If you started using Anyscale before April 2024, use Version 1.0.0 of the docs. If you're transitioning to Anyscale Preview, see the guide for how to migrate.

When using logprobs the LLM outputs the log probabilities of each output token during generation.

There are two relevant parameters for this mode:

• logprobs: default=False - When set to True the LLM outputs the log probabilities of each output token during generation.
• top_logprobs: default=None - When set to an integer value, the LLM outputs the log probabilities of the top top_logprobs most likely tokens at each token position during generation. top_logprobs must be between 0 and 5.

warning

Anyscale doesn't support meta-llama/Llama-2-70b-chat-hf and meta-llama/Llama-2-13b-chat-hf.

Example

cURL

curl "$ANYSCALE_BASE_URL/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ANYSCALE_API_KEY" \
  -d '{
    "model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
    "messages": [{"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Say 'Test'."}],
    "temperature": 0.7,
    "logprobs": true,
    "top_logprobs": 1
  }'

Python

import os
import requests

s = requests.Session()

api_base = os.getenv("ANYSCALE_BASE_URL")
token = os.getenv("ANYSCALE_API_KEY")
url = f"{api_base}/chat/completions"
body = {
    "model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
    "messages": [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Say 'Test'."}
    ],
    "temperature": 0.7,
    "logprobs": True,
    "top_logprobs": 1
}

with s.post(url, headers={"Authorization": f"Bearer {token}"}, json=body) as resp:
    print(resp.json())

OpenAI Python SDK

import openai

client = openai.OpenAI(
    base_url = "https://api.endpoints.anyscale.com/v1",
    api_key = "esecret_YOUR_API_KEY")

# Note: not all arguments are currently supported and will be ignored by the backend.
chat_completion = client.chat.completions.create(
    model="mistralai/Mixtral-8x7B-Instruct-v0.1",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Say 'Test'."}
    ],
    temperature=0.7,
    logprobs=True,
    top_logprobs=1
)
print(chat_completion.model_dump())

Example output:

{"id":"mistralai/Mixtral-8x7B-Instruct-v0.1",
 "object":"text_completion",
 "created":1705450393,
 "model":"mistralai/Mixtral-8x7B-Instruct-v0.1",
 "choices": [{"message":
                {"role":"assistant",
                 "content":"Test.",
                 "tool_calls":null,
                 "tool_call_id":null},
                 "index":0,
                 "finish_reason":"stop",
                 "logprobs":{"content":
    [
        {"token":"Test",
         "logprob":-0.12771208584308624,
         "bytes":[84,101,115,116],
         "top_logprobs": [
                {"logprob":-0.12771208584308624,
                "token":"Test",
                "bytes":[84,101,115,116]
                }
            ]
        },
        {"token":".",
         "logprob":-0.0008685392094776034,
         "bytes":[46],
         "top_logprobs": [
                {"logprob":-0.0008685392094776034,
                 "token":".",
                 "bytes":[46]
                }
            ]
        },
        {"token":"",
        "logprob":0.0,
        "bytes":[],
        "top_logprobs":[
                {"logprob":0.0,
                "token":"",
                "bytes":[]
                }
            ]
        }
    ]
                            }
            }],
"usage": {"prompt_tokens": 26,
          "completion_tokens": 3,
          "total_tokens":29
         }
}