- python==3.7.9
- torch==1.7.1
- numpy==1.19.2
- tqdm==4.54.1
- click==7.1.2
- logzero==1.6.3
- scikit-learn==0.24.1
- ruamel.yaml==0.16.12
- gdown==3.12.2
- pandas==1.1.5
- nltk==3.5
- gensim==3.8.3
- scipy==1.5.2
- transformers==4.2.1
- mlflow==1.15.0
$ pip install -r requirements.txt
If you prefer to use docker, follow below instructions.
- cuda >= 11.0
- Docker >= 19.03.5
- nvidia-docker
$ docker build -t [image_name]
Or use our pre-built image realappsdev/m2m-text
$ docker run --rm -t -d --init --name [container_name] --gpus all -v $PWD:/workspace -w /workspace --ipc=host --net=host [image_name]
$ docker exec -it [cotainer_name] zsh # Enter shell
$ ./scripts/run_model.sh
You can track experiments with mlflow.
$ ./scripts/mlflow_ui.sh # Enter http://[your-ip]:5000