CurML - GitLite

KO | EN

gitlite — search

#python #hacktoberfest #ai #llm #android #linux #docker #macos #java #windows #cli #awesome-list #security #open-source

CurML

(★ 99)

No description available.

파일 탐색기

최종 버전 다운로드 (.zip)

- - __init__.py
  - adaptive.py
  - baby_step.py
  - base.py
  - cbs.py
  - coarse_to_fine.py
  - data_parameters.py
  - dds.py
  - dihcl.py
  - lambda_step.py
  - local_to_global.py
  - meta_reweight.py
  - meta_weight_net.py
  - minimax.py
  - rl_teacher.py
  - screener_net.py
  - self_paced.py
  - superloss.py
  - transfer_teacher.py
  - utils.py
- - __init__.py
  - convnet.py
  - densenet.py
  - dla.py
  - dla_simple.py
  - dpn.py
  - efficientnet.py
  - googlenet.py
  - lenet.py
  - mobilenet.py
  - mobilenetv2.py
  - pnasnet.py
  - preact_resnet.py
  - regnet.py
  - resnet.py
  - resnext.py
  - senet.py
  - shufflenet.py
  - shufflenetv2.py
  - vgg.py
- - __init__.py
  - cifar10.py
  - cifar100.py
  - ptb.py
  - utils.py
  - wikitext.py
- - __init__.py
  - image_classifier.py
  - language_model.py
- - __init__.py
  - log.py
  - rand.py
- - flow.svg
  - framework.svg
- CurML.pdf
- demo-cl.gif
- demo-cl.mov
- adaptive.py
- baby_step+superloss.py
- baby_step.py
- base.py
- cbs.py
- coarse_to_fine.py
- data_parameters.py
- dds.py
- dihcl.py
- lambda_step.py
- local_to_global.py
- meta_reweight.py
- meta_weight_net.py
- minimax.py
- rl_teacher.py
- screener_net.py
- self_paced.py
- superloss.py
- transfer_teacher.py
- - cifar10.py
- - log.py
.gitignore
LICENSE
README.md
setup.py

# 설치 가이드

1. 코드 내려받기

git clone https://github.com/THUMNLab/CurML

깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.

cd CurML

방금 내려받은 프로젝트 폴더 안으로 이동합니다.

2. Python

쉬움 추천

사전 준비물

Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
Python 3 설치 시 'Add Python to PATH' 옵션을 꼭 체크하세요 (Windows).

pip install .

PyPI에 배포된 패키지를 바로 설치합니다. 소스 클론이 필요 없습니다.

python <실행할 파일명>.py # README에서 정확한 실행 파일명을 확인하세요

파이썬 스크립트(또는 모듈)를 실행합니다.

✅ 에러 메시지 없이 실행되고 터미널에 안내 문구가 출력되면 정상입니다.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# CurML: A Curriculum Machine Learning Library

A library & toolkit for Curriculum Learning.

The code of the ACM MM 2022 Open Source Software paper **CurML: A Curriculum Machine Learning Library**.

The paper can be downloaded from the [official website](https://github.com/THUMNLab/CurML/raw/master/docs/CurML.pdf) or in the [docs directory](https://github.com/THUMNLab/CurML/tree/master/docs).

This repository contains the first version of our curriculum learning framework.
The latest and reorganized version is provided in [CurBench](https://github.com/THUMNLab/CurBench).

## Demo
![demo](https://raw.githubusercontent.com/THUMNLab/CurML/master/docs/demo-cl.gif)

## Target

1. To reproduce, evaluate and compare existing Curriculum Learning algorithms

2. To provide baselines and benchmarks for Curriculum Learning study.

3. To utilize Curriculum Learning as a plugin to improve the performance of existing works.

## Introduction

The figure below shows the overall framework of the library. The main part is a **CL Trainer** class, composed of a **CL Algorithm** class and a **Model Trainer** class, which interact with each other through five **APIs** of the CL Algorithm class to implement the curriculum.

<img src="./docs/img/framework.svg">

An general workflow of curriculum machine learning is illustrated below.

<img src="./docs/img/flow.svg">

## Environment

1. python >= 3.6

2. pytorch >= 1.9.0

## Quick Start

``` bash
# 1. clone from the repository
git clone https://github.com/THUMNLab/CurML
cd CurML

# 2. pip install local module: curriculum
pip install -e .

# 3. run the example code
python examples/base.py
```

## Development & Extension

Every CL method implemented has been aranged in `/curriculum/algorithms`, each consists of a **Trainer class** and an **Algorithm class**. Their roles and how they interact with each other are illustrated in the figure above. `Image_classifier.py` in `/curriculum/trainer` deals with the whole learning process. It embeds the five **APIs** from existing CL methods, initializing datasets and models, then completes the whole training procedure(forwarding and back propagation). Examples in `/examples` show how you can call the implemented methods, and how to pass the hyperparameters and start training in an easy manner.

For extension, more latest CL algorithms will be added, or you can add your own CL method in `/curriculum/algorithms`, which should be consistent with the overall framework. For convenience, `/curriculum/backbones` incorporates some standard implementations of popular neural networks, including `resnet`, `vgg`, etc. For other customized datasets, you can add your dataset processing scripts in `/curriculum/datasets`, following the example in `cifar10.py`(only support `cifar10` at present). Trainers for tasks other than image classification in the field of CV, NLP, etc., will be implemented in the future.

## Cite

Please cite our [paper](https://dl.acm.org/doi/pdf/10.1145/3503161.3548549) as follows if you find our code useful:
```
@inproceedings{zhou2022curml,
  title={CurML: A Curriculum Machine Learning Library},
  author={Zhou, Yuwei and Chen, Hong and Pan, Zirui and Yan, Chuanhao and Lin, Fanqi and Wang, Xin and Zhu, Wenwu},
  booktitle={Proceedings of the 30th ACM International Conference on Multimedia},
  pages={7359--7363},
  year={2022}
}
```

You may also find our [survey paper](https://arxiv.org/pdf/2010.13166.pdf) helpful:
```
@article{wang2021survey,
  title={A survey on curriculum learning},
  author={Wang, Xin and Chen, Yudong and Zhu, Wenwu},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2021},
  publisher={IEEE}
}
```

## License
We follow [Apache license](LICENSE) across the entire codebase from v0.2.