promptmeteo.prompts package

Submodules

promptmeteo.prompts.base module

class promptmeteo.prompts.base.BasePrompt(prompt_domain: str = '', prompt_labels: str = '', prompt_detail: str = '')

Bases: ABC

Prompt class interface.

PROMPT_EXAMPLE = '\n            TEMPLATE:\n                "Here you explain the task.\n                    {__PROMPT_DOMAIN__}\n                    {__PROMPT_DETAIL__}\n\n                    {__SHOT_EXAMPLES__}\n\n                    {__PROMPT_SAMPLE__}\n                    {__CHAIN_THOUGHT__}\n                    {__ANSWER_FORMAT__}"\n\n            PROMPT_SAMPLE:\n                "Here you describe the task."\n\n            SHOT_EXAMPLES:\n                "Here you set the {__EXAMPLES__} to take into consideration."\n\n            PROMPT_DOMAIN:\n                "Here you explain the {__DOMAIN__} from the texts."\n\n            PROMPT_LABELS:\n                "Here you give the {__LABELS__} if required."\n\n            PROMPT_DETAIL:\n                "Here you can give some {__DETAIL__}"\n\n            CHAIN_THOUGHT:\n                "Explain your answer step by step."\n\n            ANSWER_FORMAT:\n                "Response just with the answer."\n        '

property domain: str

Prompt Domain.

property labels: List[str]

Prompt Labels.

classmethod read_prompt(prompt_text: str) → None

Reads a Promptmeteo prompt string to build the Task Prompt. Promptmeteo prompts are expected to follow the following template:

Parameters:

prompt_text (str) –

Returns:

Self

run() → PromptTemplate

Returns the prompt template for the current task.

property template: str

Prompt Template.

Module contents

class promptmeteo.prompts.PromptFactory

Bases: object

Factory of Prompts

static build_class(language: str, task_type: str, _model_name: str) → BasePrompt

Creates a class dinamically that inherits from BasePrompt given arguments configuracion. This new class has included its prompt from the .prompt file.

Example:

>>> PromptFactory.build_class(
>>>     language='es',
>>>     task_type='ner',
>>>     model_name='text-davinci-003')
>>> )

<class ‘__main__.Textdavinci003SpNer’>

classmethod factory_method(language: str, task_type: str, model_name: str, prompt_domain: str, prompt_labels: List[str], prompt_detail: str)

Returns and instance of a BasePrompt object depending on the task_type.

promptmeteo.prompts.get_files_taxonomy(sep: str = '_')

Convert a list of prompt files with the naming convetion of `{model_name}_{language}_{task}.prompt into a dictorionary version. Given the following folder structure:

./prompts/

├── __init__.py ├── base.py ├── google-flan-t5-small_es_classification.prompt ├── google-flan-t5-small_es_code-generation.prompt ├── google-flan-t5-small_es_ner.prompt ├── text-davinci-003_es_classification.prompt ├── text-davinci-003_es_code-generation.prompt └── text-davinci-003_es_ner.prompt

The function returns:

{‘google-flan-t5-small’: {

‘es’: {

‘classification’: {}, ‘ner’: {} }

},

‘text-davinci-003’: {

‘es’: {

‘classification’: {}, ‘code-generation’: {}, ‘ner’: {} }

}

}