A clean and scalable template to structure ML paper-code the same so that work can easily be extended and replicated.
What it does
First, install dependencies
# clone project
git clone https://github.com/YourGithubName/your-repository-name
cd your-repository-name
# [SUGGESTED] use conda environment
conda env create -f environment.yaml
conda activate lit-template
# [ALTERNATIVE] install requirements directly
pip install -r requirements.txt
Next, to obtain the main results of the paper:
# commands to get the main results
You can also run experiments with the run
script.
# fit with the demo config
./run fit --config configs/demo.yaml
# or specific command line arguments
./run fit --model MNISTModel --data MNISTDataModule --data.batch_size 32 --trainer.gpus 0
# evaluate with the checkpoint
./run test --config configs/demo.yaml --ckpt_path ckpt_path
# get the script help
./run --help
./run fit --help
@article{YourName,
title={Your Title},
author={Your team},
journal={Location},
year={Year}
}