aiconfig-extension-hugging-face 0.0.5

An extension for using Hugging Face tasks to parse models for AIConfig.

This extension contains AIConfig model parsers with two main subfolders:

1. local_inference: Loads models onto your machine and uses Hugging Face transformers and diffusors locally.
2. remote_inference_client: Uses Hugging Face's InferenceClient API to connect to models remotely

Usage

Part 1: Update and test this extension

If you are not testing an developing locally (just using the published extension), ignore this and go to Part 2

1. From the aiconfig/extensions/HuggingFace, run this command: pip3 install build && cd python && python -m build && pip3 install dist/*.whl
2. Link your local dev environment to the current dir: pip3 install -e .. Afterwards if you do pip3 list | grep aiconfig, you should see this linked to your local path. If you ever wish to use the published extension, you will need to first remove the extension: pip3 uninstall aiconfig-extension-hugging-face && pip3 install aiconfig-extension-hugging-face
3. After you're done testing, be sure to delete the generated folder(s) in the aiconfig/HuggingFace dir. It'll probably look something like python/dist and python/<package_name>.egg-info

Part 2: Importing and using this extension

pip3 install aiconfig-extension-hugging-face

1. Import the library to your code: from aiconfig_extension_hugging_face import <EXTENSION>.
2. Import the AIConfig model registery: from aiconfig import ModelRegistryParser
3. In code, add all the relevant model parser objects that you want to use from this extension to the registry. Ex: ModelParserRegistry.register_model_parser(HuggingFaceTextGenerationTransformer()). You can read the docstrings under ModelParserRegistry class for more info
4. In your AIConfig, add a field model_parsers with the model you want to use and the id of the extension you want to use: . Ex: https://github.com/lastmile-ai/aiconfig/blob/f1840995b7a12acba371a59ac3b8c69b3962fc68/cookbooks/Getting-Started/travel.aiconfig.json#L19-L22
5. Now whenever you call aiconfig.run() these model parsers will be loaded and available!