/ConvVAE

Modular and reusable Conv VAE

Primary LanguagePython

ConvVAE

Modular and reusable Conv VAE Comes with script to make images of boxes to test model Goals:

  • Modularity:

    • To use, should just have to instantiate model

    • Model object should have functionality for:

    • training ala Train(input data, target data, training time, batch size)

    • loading checkpoints ala Load(weight path)

    • saving checkpoints ala Save(weight path)

    • predicting ala Predict(input data)

    • Note this functionality is centered on data (a data translation tool may be needed later)

  • Reusability

    • Model ctor should take layers wanted and size requirements
    • Same model should be reusable for any VAE needs

Requirements:

  • Pytorch, CUDA (optional)

Usage:

  • Check VAE.py for all methods and documentation