Releases · muditbhargava66/dropgrad

What's New in Version 0.3.5?

Added support for the Lion optimizer in the ViT experiments

Implemented gradient clipping to prevent gradient explosion and improve training stability

Enhanced data augmentation techniques for better model generalization

Improved error handling and user interruption handling during training

Updated test suite to cover various aspects of DropGrad, including initialization, optimization step, drop rate scheduling, and saving of loss values

Code refactoring and documentation enhancements for better readability and maintainability

Updates in Version 3.0.0

Enhanced cross-platform compatibility: The codebase now works seamlessly on macOS, Windows, and Linux
Improved device selection logic: Automatically detects and utilizes the available hardware (MPS, CUDA, or CPU) for training
Updated dependencies: Added torchvision, torchaudio, matplotlib, and scipy as dependencies in requirements.txt and pyproject.toml
Improved visualization: Enhanced visualize.py with better plot layout and cross-platform file paths
Code cleanup and refactoring: Improved code structure and readability
Added mathematical analysis: Introduced mathematical_analysis.py to analyze the effect of DropGrad on various optimizers
Added benchmark visualizations: Introduced benchmark_visualizations.py to compare the behavior of DropGrad across optimizers and benchmarks