PromptTemplate

class openicl.PromptTemplate(template: Dict | str, column_token_map: Dict, selected_column_name: str | None = None, selected_column_map: Dict | None = None, ice_token: str | None = None)[source]
In-context Learning Prompt Template Class

This class represents a template that guides the generation of prompts in the retrieval or inference process.

template

A custom template dictionary or string. If a dictionary, the keys of the dictionary represent the values of the output_column, and the values represent the corresponding generated statement. If a string, it represents a string template.

Type:

Dict or str

column_token_map

A dictionary mapping column names to specific tokens. The tokens will be replaced by data in the corresponding column (one piece each time) during the retrieval or inference process.

Type:

Dict

selected_column_name

Used only with string-type templates. A specific column that needs its value to be mapped.

Type:

str, optional

selected_column_map

Used only with string-type templates. Maps the value of the column selected_column_name.

Type:

Dict, optional

ice_token

A string that represents the specific token mapping from in-context examples. None if you want to use this template only to generate in-context examples, otherwise it can be used to generate the final prompt that is fed into the PLM. The ice_token will be invisible when generating in-context examples.

Type:

str, optional

generate_ice_item(entry: Dict, label: Hashable) → str[source]

Generate in-context example based on the provided entry data.

Parameters:

  • entry (Dict) – A piece of data to be used for generating the in-context example.

  • label (Hashable) – The value of the output field.

Returns:

The generated in-context example.

Return type:

str

generate_item(entry: Dict, output_field: Hashable | None = None, output_field_replace_token: str | None = '', ice_field_replace_token: str | None = '') → str[source]

Generate an item based on the provided entry data, as well as optional output field and ice field tokens.

Parameters:

  • entry (Dict) – A piece of data.

  • output_field (Hashable, optional) – Column name of output field. Defaults to None.

  • output_field_replace_token (str, optional) – Tokens used to replace output field. Defaults to ''.

  • ice_field_replace_token (str, optional) – Tokens used to replace the ice_token. Defaults to ''.

Returns:

The generated item.

Return type:

str

generate_label_prompt_item(entry: Dict, ice: str, label: Hashable) → str[source]

Generate prompt based on entry data, ice in-context example, and the corresponding label.

Parameters:

  • entry (Dict) – A piece of data containing the input field content.

  • ice (str) – The generated in-context example.

  • label (Hashable) – The value of the output field.

Raises:

ValueError – If the ice_token attribute of the PromptTemplate instance is None.

Returns:

The generated prompt.

Return type:

str