CoolGraph - GitLite

KO | EN

gitlite — search

#javascript #python #ai #llm #golang #go #machine-learning #macos #nodejs #java #cli #awesome-list #security #ios

CoolGraph

(★ 134)

Make GNN easy to start with

파일 탐색기

최종 버전 다운로드 (.zip)

- - main.yml
  - python-publish.yml
- - __init__.py
  - config.py
  - run.py
- - full.yaml
  - in_memory_data.yaml
  - in_memory_data2.yaml
- - __init__.py
  - batch.py
  - data_processor.py
  - dataset.py
  - loaders.py
  - utils.py
- - .gitkeep
  - __init__.py
  - antifraud.py
  - multitarget.py
  - nonhomophilous.py
  - ogbn_proteins.py
  - s_ffsd.py
- - __init__.py
  - mlflow_logging.py
- - __init__.py
  - categorical_embeddings.py
  - graphconv_gnn.py
  - nnconv_gnn.py
- - __init__.py
  - example_objective.py
- - __init__.py
  - helpers.py
  - loss.py
  - metrics.py
  - plot.py
  - trainer.py
- __init__.py
- runners.py
- - Penn94.mat
- - AntiFraud.ipynb
  - NonHomophilous.ipynb
  - S_FFSD.ipynb
- - image_2024-08-21_13-59-13.png
- benchmarks.ipynb
- categorical_features_usage_examples.ipynb
- CoolGraph_usage_examples.ipynb
- datasets_test.ipynb
- Easy_start.ipynb
- graph_features_usage_examples.ipynb
- How_to_work_with_configs.ipynb
- How_to_work_with_datasets.ipynb
- Indices_for_DataLoader.ipynb
- Integration_with_mlflow.ipynb
- predict_proba_examples.ipynb
- Usage_examples.ipynb
- Visualization example.ipynb
- YelpChi_dataset_with_edge_attr.ipynb
- - full.yaml
- - edges.parquet
  - labels.parquet
  - nodes.parquet
- test_get_auto_batch_size.py
- test_run.py
- test_run_config.py
.gitignore
CONTRIBUTING.md
environment.yml
LICENSE.md
Makefile
MANIFEST.in
README.md
setup.py

# 설치 가이드

1. 코드 내려받기

git clone https://github.com/MTSWebServices/CoolGraph

깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.

cd CoolGraph

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.

jupyter notebook

브라우저에서 노트북(.ipynb) 파일들을 열람하고 실행할 수 있는 Jupyter 화면을 켭니다.

✅ 에러 메시지 없이 실행되고 터미널에 안내 문구가 출력되면 정상입니다.

3. Make

보통

사전 준비물

Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
Make Linux/macOS는 보통 기본 설치되어 있습니다. Windows는 별도 설치(예: MSYS2, WSL)가 필요합니다.

make

생성된 빌드 설정을 바탕으로 실제 컴파일을 진행해 실행 파일을 만듭니다.

✅ 에러 없이 끝나면 성공입니다. 생성된 실행 파일을 직접 실행해보세요.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

CoolGraph is an easy-to-use Python library using Graph Neural Networks for node classification. The CoolGraph contains several architectures that will help you train a network using two lines of code.

Thus, the parameters for training have already been selected and collected in configs, but you can change them as you wish.

Also, if for some reason the selected parameters do not ok for you, it is possible to use the search for hyperparameters with Optuna.

Moreover, your experiments can be saved in Mlflow and fully tracked.

All you need is graph-structured data.

# Main advantages of CoolGraph:
 - quick start in one line of code
 - good quality of base models, comparable to state of the art
 - the best model architecture automatic search via optuna
 - heterogeneous graph support
 - multitarget support
 - both node and edge features support
 - categorical and graph feaures usage
 - tracking experiments with mlflow

# Benchmarks

| *Dataset*            | *Main Metric* | *Train* | *Valid* | *Test*  | *Runner Score* | *HypeRunner Score* | *SOTA* | *Rank* |
|----------------------|---------------|---------|---------|---------|----------------|--------------------|--------|--------|
| **Amazon Computers** | accuracy      | 0.6     | 0.2     | 0.2     | 0.918          | 0.915              | 0.939  | 3      |
| **Amazon Photo**     | accuracy      | 0.6     | 0.2     | 0.2     | 0.959          | 0.961              | 0.967  | 3      |
| **Amzon-Fraud**      | roc-auc       | 0.4     | 0.2     | 0.4     | 0.956          | 0.960              | 0.975  | 4      |
| **YelpChi**          | roc-auc       | 0.4     | 0.2     | 0.4    | 0.856           | 0.890              | 0.950  | 5      |
| **Multitarget 10k**  | roc-auc       | 0.6     | 0.2     | 0.2    | 0.730           | 0.838              |        |        |
| **Multitarget 50k**  | roc-auc       | 0.6     | 0.2     | 0.2    | 0.756           | 0.841              |        |        |
| **Penn94**           | accuracy      | 0.6     | 0.2     | 0.2    | 0.791           | 0.829              | 0.861  | 12     |
| **Genius**           | roc-auc       | 0.6     | 0.2     | 0.2    | 0.902           | 0.902              | 0.915  | 12     |

# Installation
## Install with creating conda env

`conda deactivate` <br>
`conda create -n cool_graph_env python=3.8 cudatoolkit=11.3.1 pytorch=1.12.0=py3.8_cuda11.3_cudnn8.3.2_0 cxx-compiler=1.5.1 pyg=2.2.0=py38_torch_1.12.0_cu113 pyarrow=11.0.0 numpy=1.23.5 pandas=1.4.4 pip=22.3.1 py=1.11.0 mysqlclient=2.0.3 sqlite=3.38.2 psycopg2=2.8.6 optuna=2.10.1 -c nvidia -c pytorch -c conda-forge -c pyg`  <br>
`conda activate cool_graph_env`.   <br>
`pip install cool-graph`

## Install with creating conda env from yml file

`conda deactivate` <br>
`conda env create -f environment.yml`  <br>
`conda activate cool_graph_env`  <br>
`pip install cool-graph`

# Get started

## Basic usage

```python
# Load Dataset
from torch_geometric import datasets
data = datasets.Amazon(root='./data/Amazon', name='Computers').data

# Train GNN model
from cool_graph.runners import Runner
runner = Runner(data)
result = runner.run()
```

You can override default parameters and/or read parameters from config file
```python
runner = Runner(data, 
                metrics=['accuracy', ...], 
                batch_size='auto', 
                train_size=0.7, 
                test_size=0.3, 
                overrides=['training.n_epochs=1', ...], 
                config_path=...)
result = runner.run()         
print(result['best_loss'])       
```
## Runner with Optuna
You can run HypeRunner for the best GNN architecture search
```python
runner = HypeRunner(data)
result = runner.optimize_run()

```
For more information look at Tutorials

# Tutorials

1. [Easy start](/notebooks/Easy_start.ipynb) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1fyjfK5pZTtAz5axZydFb59eIQdKrBbR8)
2. [Main features](/notebooks/Usage_examples.ipynb) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/168saYLMPRXc7BOELzVKG0049en4Rr3Jx)
3. [Making predictions and metric calculation](h/notebooks/predict_proba_examples.ipynb)
4. [Working with categorical features](/notebooks/categorical_features_usage_examples.ipynb)
5. [Working with graph features](/notebooks/graph_features_usage_examples.ipynb)
5. [Creating your own data loaders](/notebooks/Indices_for_DataLoader.ipynb)
6. [Working with configs](/notebooks/How_to_work_with_configs.ipynb)
7. [Benchmarks](/notebooks/benchmarks.ipynb)
8. [Integration with Mlflow](/notebooks/Integration_with_mlflow.ipynb)

## Library Structure

The directory structure cool_graph:

```
├── config                       <- Config structure
│   ├── data                     <- Data configs (Datasets on disk)
│   ├── logging                  <- MLFlow configs
│   ├── metrics                  <- Metrics configs
│   ├── model_params             <- NN model parameters configs
│   ├── training                 <- Training  configs
│   ├── full.yaml                <- Main config for CLI 
│   ├── in_memory_data.yaml      <- Main config for notebook (GraphConv)
│   ├── in_memory_data2.yaml     <- Main config for notebook (NNConv)
│
├── cli                    <- Cli commands
├── data                   <- Data processing, data loaders, batch sizes
├── datasets               <- CoolGraph datasets
├── logging                <- MLflow logging experiments
├── models                 <- CoolGraph models
├── parameter_search       <- Sampling model params for Optuna
├── train                  <- Trainin / eval / metrics code
├── runners.py             <- Run training (CLI + Notebook)

```

## Development

**add dev dependencies here (or switch to poetry)**

Install the package using one of the options described above adding `-e` flag to `pip install`.

Run `git checkout -b <new_branch_name>`.

Introduce changes to the branch.

Test your changes using `make test`.

Add tests if necessary (e.g. coverage fell below 80%), run `make test` again.

Ensure that codestyle is OK with `make verify_format`.

If not, run `make format`.

After that commit -> push -> PR.

## Authors
 
* [Diana Pavlikova](https://github.com/dapavlik)
* [Sergey Kuliev](https://github.com/kuliev-sd)
* [Igor Inozemtsev](https://github.com/inozemtsev)
* [Nikita Zelinskiy](https://github.com/nikita-ds)
* [Alexey Pristaiko](https://github.com/qwertd105)
* [Vladislav Kuznetsov](https://github.com/AnanasClassic)

## License

`The MIT License (MIT)
Copyright 2023 MTS (Mobile Telesystems). All rights reserved.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.`