Pytorch-Lite

Easy to use small framework for faster model development and visualization

Usage

• clone the repo
• create conda environment (conda env create -f env_specs.yml)
• choose or define a model (models/your_model.py)
• choose or define a dataset_loader
• modify constants.py
• run model
• run tensorboard (tensorboard --logdir=runs) to see the results

Future works

1. Refactor and improve backbone functionality
2. Add examples (MixNet, EfficientNet, gans etc) and improve documentation
3. Add multiple models and layers
4. Add more image enhancement techniques
5. Add or facilitate more complex image visualization methods (matplotlib ... facets)
6. Add python "artisan" for easier use