Allen's PyTorch Codebook

• I use this repo as my PyTorch training pipeline template.
• Trying to implement the model I know using PyTorch.
• Finally make this repo as a template of Music Generation VAE Project.

TODO

Architecture

...

Models

2. MNIST Classifier.
3. Circle AutoEncoder.
   • (x, y) -> Enc -> z{2} -> Dec -> (x, y)
4. MNIST AutoEncoder.

Environment

conda create torch python=3.8

pip install -r requirements.txt
# pip install jupyter

Run example

python main.py --config circle.yaml

Run in Docker

• I use pytorch/pytorch:1.12.0-cuda11.3-cudnn8-runtime as the source image.
• In the future I will implement wandb so that you can visialize every just like running example locally.

Build

• Remember to set your $MY_WANDB_API.

docker build --build-arg WANDB_API=$MY_WANDB_API -t torch-codebook . --no-cache

Run

# Run bash
docker run --gpus all -it --rm torch-codebook bash

# You can run training command like this
docker run --gpus all -it --rm torch-codebook python main.py --config circle.yaml --gpu_id cuda

Workflow

1. Create a config yaml file use in workflow.
2. Create dataset.
3. Create model.
4. Create Solver.
5. Create your main.py to use the solver.