Hugging Face's logo

Transformers documentation

Agents & Tools

You are viewing v4.30.0 version. A newer version v4.47.1 is available.
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Agents & Tools

Transformers Agent is an experimental API which is subject to change at any time. Results returned by the agents can vary as the APIs or underlying models are prone to change.

To learn more about agents and tools make sure to read the introductory guide. This page contains the API docs for the underlying classes.

Agents

We provide three types of agents: HfAgent uses inference endpoints for opensource models, LocalAgent uses a model of your choice locally and OpenAiAgent uses OpenAI closed models.

HfAgent

class transformers.HfAgent

< >

( url_endpoint token = None chat_prompt_template = None run_prompt_template = None additional_tools = None )

Parameters

  • url_endpoint (str) — The name of the url endpoint to use.
  • token (str, optional) — The token to use as HTTP bearer authorization for remote files. If unset, will use the token generated when running huggingface-cli login (stored in ~/.huggingface).
  • chat_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the chat method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named chat_prompt_template.txt in this repo in this case.
  • run_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the run method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named run_prompt_template.txt in this repo in this case.
  • additional_tools (Tool, list of tools or dictionary with tool values, optional) — Any additional tools to include on top of the default ones. If you pass along a tool with the same name as one of the default tools, that default tool will be overridden.

Agent that uses an inference endpoint to generate code.

Example:

from transformers import HfAgent

agent = HfAgent("https://api-inference.huggingface.co/models/bigcode/starcoder")
agent.run("Is the following `text` (in Spanish) positive or negative?", text="¡Este es un API muy agradable!")

LocalAgent

class transformers.LocalAgent

< >

( model tokenizer chat_prompt_template = None run_prompt_template = None additional_tools = None )

Parameters

  • model (PreTrainedModel) — The model to use for the agent.
  • tokenizer (PreTrainedTokenizer) — The tokenizer to use for the agent.
  • chat_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the chat method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named chat_prompt_template.txt in this repo in this case.
  • run_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the run method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named run_prompt_template.txt in this repo in this case.
  • additional_tools (Tool, list of tools or dictionary with tool values, optional) — Any additional tools to include on top of the default ones. If you pass along a tool with the same name as one of the default tools, that default tool will be overridden.

Agent that uses a local model and tokenizer to generate code.

Example:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, LocalAgent

checkpoint = "bigcode/starcoder"
model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto", torch_dtype=torch.bfloat16)
tokenizer = AutoTokenizer.from_pretrained(checkpoint)

agent = LocalAgent(model, tokenizer)
agent.run("Draw me a picture of rivers and lakes.")

from_pretrained

< >

( pretrained_model_name_or_path **kwargs )

Parameters

  • pretrained_model_name_or_path (str or os.PathLike) — The name of a repo on the Hub or a local path to a folder containing both model and tokenizer. kwargs — Keyword arguments passed along to from_pretrained().

Convenience method to build a LocalAgent from a pretrained checkpoint.

Example:

import torch
from transformers import LocalAgent

agent = LocalAgent.from_pretrained("bigcode/starcoder", device_map="auto", torch_dtype=torch.bfloat16)
agent.run("Draw me a picture of rivers and lakes.")

OpenAiAgent

class transformers.OpenAiAgent

< >

( model = 'text-davinci-003' api_key = None chat_prompt_template = None run_prompt_template = None additional_tools = None )

Parameters

  • model (str, optional, defaults to "text-davinci-003") — The name of the OpenAI model to use.
  • api_key (str, optional) — The API key to use. If unset, will look for the environment variable "OPENAI_API_KEY".
  • chat_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the chat method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named chat_prompt_template.txt in this repo in this case.
  • run_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the run method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named run_prompt_template.txt in this repo in this case.
  • additional_tools (Tool, list of tools or dictionary with tool values, optional) — Any additional tools to include on top of the default ones. If you pass along a tool with the same name as one of the default tools, that default tool will be overridden.

Agent that uses the openai API to generate code.

The openAI models are used in generation mode, so even for the chat() API, it’s better to use models like "text-davinci-003" over the chat-GPT variant. Proper support for chat-GPT models will come in a next version.

Example:

from transformers import OpenAiAgent

agent = OpenAiAgent(model="text-davinci-003", api_key=xxx)
agent.run("Is the following `text` (in Spanish) positive or negative?", text="¡Este es un API muy agradable!")

AzureOpenAiAgent

class transformers.AzureOpenAiAgent

< >

( deployment_id api_key = None resource_name = None api_version = '2022-12-01' is_chat_model = None chat_prompt_template = None run_prompt_template = None additional_tools = None )

Parameters

  • deployment_id (str) — The name of the deployed Azure openAI model to use.
  • api_key (str, optional) — The API key to use. If unset, will look for the environment variable "AZURE_OPENAI_API_KEY".
  • resource_name (str, optional) — The name of your Azure OpenAI Resource. If unset, will look for the environment variable "AZURE_OPENAI_RESOURCE_NAME".
  • api_version (str, optional, default to "2022-12-01") — The API version to use for this agent.
  • is_chat_mode (bool, optional) — Whether you are using a completion model or a chat model (see note above, chat models won’t be as efficient). Will default to gpt being in the deployment_id or not.
  • chat_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the chat method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named chat_prompt_template.txt in this repo in this case.
  • run_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the run method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named run_prompt_template.txt in this repo in this case.
  • additional_tools (Tool, list of tools or dictionary with tool values, optional) — Any additional tools to include on top of the default ones. If you pass along a tool with the same name as one of the default tools, that default tool will be overridden.

Agent that uses Azure OpenAI to generate code. See the official documentation to learn how to deploy an openAI model on Azure

The openAI models are used in generation mode, so even for the chat() API, it’s better to use models like "text-davinci-003" over the chat-GPT variant. Proper support for chat-GPT models will come in a next version.

Example:

from transformers import AzureOpenAiAgent

agent = AzureAiAgent(deployment_id="Davinci-003", api_key=xxx, resource_name=yyy)
agent.run("Is the following `text` (in Spanish) positive or negative?", text="¡Este es un API muy agradable!")

Agent

class transformers.Agent

< >

( chat_prompt_template = None run_prompt_template = None additional_tools = None )

Parameters

  • chat_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the chat method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named chat_prompt_template.txt in this repo in this case.
  • run_prompt_template (str, optional) — Pass along your own prompt if you want to override the default template for the run method. Can be the actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named run_prompt_template.txt in this repo in this case.
  • additional_tools (Tool, list of tools or dictionary with tool values, optional) — Any additional tools to include on top of the default ones. If you pass along a tool with the same name as one of the default tools, that default tool will be overridden.

Base class for all agents which contains the main API methods.

chat

< >

( task return_code = False remote = False **kwargs )

Parameters

  • task (str) — The task to perform
  • return_code (bool, optional, defaults to False) — Whether to just return code and not evaluate it.
  • remote (bool, optional, defaults to False) — Whether or not to use remote tools (inference endpoints) instead of local ones.
  • kwargs (additional keyword arguments, optional) — Any keyword argument to send to the agent when evaluating the code.

Sends a new request to the agent in a chat. Will use the previous ones in its history.

Example:

from transformers import HfAgent

agent = HfAgent("https://api-inference.huggingface.co/models/bigcode/starcoder")
agent.chat("Draw me a picture of rivers and lakes")

agent.chat("Transform the picture so that there is a rock in there")

run

< >

( task return_code = False remote = False **kwargs )

Parameters

  • task (str) — The task to perform
  • return_code (bool, optional, defaults to False) — Whether to just return code and not evaluate it.
  • remote (bool, optional, defaults to False) — Whether or not to use remote tools (inference endpoints) instead of local ones.
  • kwargs (additional keyword arguments, optional) — Any keyword argument to send to the agent when evaluating the code.

Sends a request to the agent.

Example:

from transformers import HfAgent

agent = HfAgent("https://api-inference.huggingface.co/models/bigcode/starcoder")
agent.run("Draw me a picture of rivers and lakes")

prepare_for_new_chat

< >

( )

Clears the history of prior calls to chat().

Tools

load_tool

transformers.load_tool

< >

( task_or_repo_id model_repo_id = None remote = False token = None **kwargs )

Parameters

  • task_or_repo_id (str) — The task for which to load the tool or a repo ID of a tool on the Hub. Tasks implemented in Transformers are:

    • "document-question-answering"
    • "image-captioning"
    • "image-question-answering"
    • "image-segmentation"
    • "speech-to-text"
    • "summarization"
    • "text-classification"
    • "text-question-answering"
    • "text-to-speech"
    • "translation"
  • model_repo_id (str, optional) — Use this argument to use a different model than the default one for the tool you selected.
  • remote (bool, optional, defaults to False) — Whether to use your tool by downloading the model or (if it is available) with an inference endpoint.
  • token (str, optional) — The token to identify you on hf.co. If unset, will use the token generated when running huggingface-cli login (stored in ~/.huggingface).
  • kwargs (additional keyword arguments, optional) — Additional keyword arguments that will be split in two: all arguments relevant to the Hub (such as cache_dir, revision, subfolder) will be used when downloading the files for your tool, and the others will be passed along to its init.

Main function to quickly load a tool, be it on the Hub or in the Transformers library.

Tool

class transformers.Tool

< >

( *args **kwargs )

A base class for the functions used by the agent. Subclass this and implement the __call__ method as well as the following class attributes:

  • description (str) — A short description of what your tool does, the inputs it expects and the output(s) it will return. For instance ‘This is a tool that downloads a file from a url. It takes the url as input, and returns the text contained in the file’.
  • name (str) — A performative name that will be used for your tool in the prompt to the agent. For instance "text-classifier" or "image_generator".
  • inputs (List[str]) — The list of modalities expected for the inputs (in the same order as in the call). Modalitiies should be "text", "image" or "audio". This is only used by launch_gradio_demo or to make a nice space from your tool.
  • outputs (List[str]) — The list of modalities returned but the tool (in the same order as the return of the call method). Modalitiies should be "text", "image" or "audio". This is only used by launch_gradio_demo or to make a nice space from your tool.

You can also override the method setup() if your tool as an expensive operation to perform before being usable (such as loading a model). setup() will be called the first time you use your tool, but not at instantiation.

from_gradio

< >

( gradio_tool )

Creates a Tool from a gradio tool.

from_hub

< >

( repo_id: str model_repo_id: typing.Optional[str] = None token: typing.Optional[str] = None remote: bool = False **kwargs )

Parameters

  • repo_id (str) — The name of the repo on the Hub where your tool is defined.
  • model_repo_id (str, optional) — If your tool uses a model and you want to use a different model than the default, you can pass a second repo ID or an endpoint url to this argument.
  • token (str, optional) — The token to identify you on hf.co. If unset, will use the token generated when running huggingface-cli login (stored in ~/.huggingface).
  • remote (bool, optional, defaults to False) — Whether to use your tool by downloading the model or (if it is available) with an inference endpoint.
  • kwargs (additional keyword arguments, optional) — Additional keyword arguments that will be split in two: all arguments relevant to the Hub (such as cache_dir, revision, subfolder) will be used when downloading the files for your tool, and the others will be passed along to its init.

Loads a tool defined on the Hub.

push_to_hub

< >

( repo_id: str commit_message: str = 'Upload tool' private: typing.Optional[bool] = None token: typing.Union[bool, str, NoneType] = None create_pr: bool = False )

Parameters

  • repo_id (str) — The name of the repository you want to push your tool to. It should contain your organization name when pushing to a given organization.
  • commit_message (str, optional, defaults to "Upload tool") — Message to commit while pushing.
  • private (bool, optional) — Whether or not the repository created should be private.
  • token (bool or str, optional) — The token to use as HTTP bearer authorization for remote files. If unset, will use the token generated when running huggingface-cli login (stored in ~/.huggingface).
  • create_pr (bool, optional, defaults to False) — Whether or not to create a PR with the uploaded files or directly commit.

Upload the tool to the Hub.

save

< >

( output_dir )

Parameters

  • output_dir (str) — The folder in which you want to save your tool.

Saves the relevant code files for your tool so it can be pushed to the Hub. This will copy the code of your tool in output_dir as well as autogenerate:

  • a config file named tool_config.json
  • an app.py file so that your tool can be converted to a space
  • a requirements.txt containing the names of the module used by your tool (as detected when inspecting its code)

You should only use this method to save tools that are defined in a separate module (not __main__).

setup

< >

( )

Overwrite this method here for any operation that is expensive and needs to be executed before you start using your tool. Such as loading a big model.

PipelineTool

class transformers.PipelineTool

< >

( model = None pre_processor = None post_processor = None device = None device_map = None model_kwargs = None token = None **hub_kwargs )

Parameters

  • model (str or PreTrainedModel, optional) — The name of the checkpoint to use for the model, or the instantiated model. If unset, will default to the value of the class attribute default_checkpoint.
  • pre_processor (str or Any, optional) — The name of the checkpoint to use for the pre-processor, or the instantiated pre-processor (can be a tokenizer, an image processor, a feature extractor or a processor). Will default to the value of model if unset.
  • post_processor (str or Any, optional) — The name of the checkpoint to use for the post-processor, or the instantiated pre-processor (can be a tokenizer, an image processor, a feature extractor or a processor). Will default to the pre_processor if unset.
  • device (int, str or torch.device, optional) — The device on which to execute the model. Will default to any accelerator available (GPU, MPS etc…), the CPU otherwise.
  • device_map (str or dict, optional) — If passed along, will be used to instantiate the model.
  • model_kwargs (dict, optional) — Any keyword argument to send to the model instantiation.
  • token (str, optional) — The token to use as HTTP bearer authorization for remote files. If unset, will use the token generated when running huggingface-cli login (stored in ~/.huggingface).
  • hub_kwargs (additional keyword arguments, optional) — Any additional keyword argument to send to the methods that will load the data from the Hub.

A Tool tailored towards Transformer models. On top of the class attributes of the base class Tool, you will need to specify:

  • model_class (type) — The class to use to load the model in this tool.
  • default_checkpoint (str) — The default checkpoint that should be used when the user doesn’t specify one.
  • pre_processor_class (type, optional, defaults to AutoProcessor) — The class to use to load the pre-processor
  • post_processor_class (type, optional, defaults to AutoProcessor) — The class to use to load the post-processor (when different from the pre-processor).

decode

< >

( outputs )

Uses the post_processor to decode the model output.

encode

< >

( raw_inputs )

Uses the pre_processor to prepare the inputs for the model.

forward

< >

( inputs )

Sends the inputs through the model.

setup

< >

( )

Instantiates the pre_processor, model and post_processor if necessary.

RemoteTool

class transformers.RemoteTool

< >

( endpoint_url = None token = None tool_class = None )

Parameters

  • endpoint_url (str) — The url of the endpoint to use.
  • token (str, optional) — The token to use as HTTP bearer authorization for remote files. If unset, will use the token generated when running huggingface-cli login (stored in ~/.huggingface).
  • tool_class (type, optional) — The corresponding tool_class if this is a remote version of an existing tool. Will help determine when the output should be converted to another type (like images).

A Tool that will make requests to an inference endpoint.

extract_outputs

< >

( outputs )

You can override this method in your custom class of RemoteTool to apply some custom post-processing of the outputs of the endpoint.

prepare_inputs

< >

( *args **kwargs )

Prepare the inputs received for the HTTP client sending data to the endpoint. Positional arguments will be matched with the signature of the tool_class if it was provided at instantation. Images will be encoded into bytes.

You can override this method in your custom class of RemoteTool.

launch_gradio_demo

transformers.launch_gradio_demo

< >

( tool_class: Tool )

Parameters

  • tool_class (type) — The class of the tool for which to launch the demo.

Launches a gradio demo for a tool. The corresponding tool class needs to properly implement the class attributes inputs and outputs.

←Pipelines for webserver inference Auto Classes→