mlflow.llm

The mlflow.llm module provides utilities for Large Language Models (LLMs).

mlflow.llm.log_predictions(inputs: List[Union[str, Dict[str, str]]], outputs: List[str], prompts: List[Union[str, Dict[str, str]]]) → None

Note

Experimental: This function may change or be removed in a future release without warning.

Log a batch of inputs, outputs and prompts for the current evaluation run. If no run is active, this method will create a new active run.

Parameters

• inputs – Union of either List of input strings or List of input dictionary

• outputs – List of output strings

• prompts – Union of either List of prompt strings or List of prompt dictionary

Returns

None

Example

import mlflow

inputs = [
    {
        "question": "How do I create a Databricks cluster with UC access?",
        "context": "Databricks clusters are ...",
    },
]

outputs = [
    "<Instructions for cluster creation with UC enabled>",
]

prompts = [
    "Get Databricks documentation to answer all the questions: {input}",
]


with mlflow.start_run():
    # Log llm predictions
    mlflow.llm.log_predictions(inputs, outputs, prompts)