maxframe.learn.contrib.llm.text.generate

maxframe.learn.contrib.llm.text.generate(data, model: TextLLM, prompt_template: List[Dict[str, Any]], params: Dict[str, Any] = None)

Generate text using a text language model based on given data and prompt template.

Parameters:

• data (DataFrame or Series) – Input data used for generation. Can be maxframe DataFrame, Series that contain text to be processed.

• model (TextLLM) – Language model instance used for text generation.

• prompt_template (List[Dict[str, str]]) –

  Dictionary containing the conversation messages template. Use {col_name} as a placeholder to reference column data from input data.

  Usually in format of [{“role”: “user”, “content”: “{query}”}], same with openai api schema.

• params (Dict[str, Any], optional) – Additional parameters for generation configuration, by default None. Can include settings like temperature, max_tokens, etc.

Returns:

Generated text raw response and success status. If the success is False, the generated text will return the error message.

Return type:

DataFrame

Examples

>>> from maxframe.learn.contrib.llm.models.managed import ManagedTextLLM
>>> import maxframe.dataframe as md
>>>
>>> # Initialize the model
>>> llm = ManagedTextLLM(name="Qwen2.5-0.5B-instruct")
>>>
>>> # Prepare prompt template
>>> messages = [
...     {
...         "role": "user",
...         "content": "Help answer following question: {query}",
...     },
... ]

>>> # Create sample data
>>> df = md.DataFrame({"query": ["What is machine learning?"]})
>>>
>>> # Generate response
>>> result = generate(df, llm, prompt_template=messages)
>>> result.execute()