Skip to main content

PaddleFSL is a Python library for few-shot learning

Project description

PaddleFSL

PaddleFSL is a Python library for few-shot learning (FSL) built upon PaddlePaddle 2.0.

  • Provide various FSL solutions which are applicable to diverse applications.
  • Contain detailed annotations and tutorial examples, such that users can easily develop and compare different FSL solutions.
  • Can be easily deployed on various training platforms.

Overview

  • paddlefsl: The proposed package for FSL.
    • paddlefsl.backbones: It provides representation learning models to extract sample representation, such as CNN, GNN and PLM.
    • paddlefsl.datasets: It provides publicly accessible benchmark few-shot datasets of diverse application scenarios, such as Omniglot, FewRel, FewGLUE and Tox21.
    • paddlefsl.model_zoo: It provides classical FSL models which particularly deal with the lack of labeled data, such as ProtoNet, MAML and PET.
    • paddlefsl.task_sampler: It provides APIs to split datasets into the typical N-way K-shot tasks.
    • paddlefsl.utils: It provides auxiliary APIs of PaddleFSL.
  • examples: It provides examples of applying PaddleFSL to diverse application scenarios, such as computer vision tasks, natural language tasks and bioinformatics tasks.
  • raw_data: It provides guides to download and place the raw data files.
  • test: It provides unit test files of functions and classes.

Installation

To use our package, users should first install paddlepaddle-v2.0.0 or later versions, see https://www.paddlepaddle.org.cn/install .

First, install our required packages.

# Clone our repository.
git clone https://github.com/tata1661/FSL-Mate.git
# Install requirements.
cd FSL-Mate/PaddleFSL
pip install -r requirements.txt

Second, update environment variables.

# Please edit env.sh to set the correct path of FSL-paddletoolkit directory.
# Then do:
source env.sh
# If you want to use our package frequently, you can add environment variables into .bashrc
cat env.sh >> ~/.bashrc
source ~/.bashrc

Finally, check whether the installation is successful.

# Start a python interpreter
python
>>> import paddlefsl
>>> paddlefsl.__version__
'1.1.0'

Contributing

PaddleFSL is mainly contributed by W Group led by Yaqing Wang. The full list of PaddleFSL contributors is here.

We also welcome and appreciate community contribution to improve PaddleFSL, such as introducing new datasets, models and algorithms, designing new features and fixing bugs. The codes can be contributed via pull requests and code review. Please also feel free to open an issue for feedbacks or advices.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

paddlefsl-1.1.0.tar.gz (57.2 kB view details)

Uploaded Source

Built Distribution

paddlefsl-1.1.0-py3-none-any.whl (101.0 kB view details)

Uploaded Python 3

File details

Details for the file paddlefsl-1.1.0.tar.gz.

File metadata

  • Download URL: paddlefsl-1.1.0.tar.gz
  • Upload date:
  • Size: 57.2 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.0 CPython/3.7.6

File hashes

Hashes for paddlefsl-1.1.0.tar.gz
Algorithm Hash digest
SHA256 745647cfb24ae25155c6882d0ca6e1e89f0729de0fd527b95137a833aa758db9
MD5 9dc8bbf04d1a5c583bd4a90a275928d2
BLAKE2b-256 7afb30d4d6c5ed67ae8b4097458d1131264b1dd5f2387f01019eaa3bf725a5de

See more details on using hashes here.

File details

Details for the file paddlefsl-1.1.0-py3-none-any.whl.

File metadata

  • Download URL: paddlefsl-1.1.0-py3-none-any.whl
  • Upload date:
  • Size: 101.0 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.0 CPython/3.7.6

File hashes

Hashes for paddlefsl-1.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 22e9304f6eafd2523538b9cc320790e9d43724d0b5381176351552b845641978
MD5 f244a7afb4f9420852150a66389e86cf
BLAKE2b-256 fb4a25d1959a8f1fe5ee400f32fc9fc8b56d4fd6fc25315e23c0171f6e705e2a

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page