- (multi-node/gpu) (training/validation/test)
- Weight and bias (Wandb) MLops
- Easy Code Structure
- CIFAR10/100 dataset, (*you can add your custom dataset easily)
- Customizable configuration
- Install Anaconda
wget https://repo.anaconda.com/archive/Anaconda3-2021.05-Linux-x86_64.sh
bash ~/Desktop/<anacondafile> -b -p
source anaconda3/bin/activate
- Create Environment
conda create -n torch python=3.9
conda activate torch
conda install pytorch==1.12.0 torchvision==0.13.0 -c pytorch
pip install tqdm easydict wandb imageio
cd canonicalSGD
python setup.py develop
- WANDB Enter this link
- create your account (you can use your Github account)
- in
config/*****.yaml
edit wandb setting - run our script and type certification code.
- Done
** You can turn off the wandb log by editing the below line**
os.environ['WANDB_SILENT'] = "true"
in agents/mnist.py
python tools/main.py --config config/...yaml --resume (0 or 1)
python tools/main.py --config config/...yaml --test 1
sh tools/dist_train.sh <ConfigFile> <#GPUS> <1(optional for resume)>
sh tools/dist_test.sh <ConfigFile> <#GPUS>
- CIFAR100 dataset: 77%
This project is licensed under MIT License - see the LICENSE file for details