Skip to content

Gradient Cache v1.0.0 - First Release

Latest

Choose a tag to compare

@JonSnow1807 JonSnow1807 released this 06 Aug 03:47
· 2 commits to main since this release

Gradient Cache v1.0.0 Release Notes

🎉 First Public Release

Features

  • 90%+ GPU memory savings during neural network training
  • 100x gradient compression with minimal accuracy impact
  • Simple 3-line integration with any PyTorch model
  • Support for Metaflow and PyTorch Lightning
  • Tested on NVIDIA A100 and T4 GPUs

Performance

  • Compression ratio: 100x (keeps top 1% of gradients)
  • Memory savings: 90-95% of gradient memory
  • Training overhead: ~10-15%
  • Enables 2-3x larger batch sizes

Files Included

  • gradient_cache-1.0.0-py3-none-any.whl - Pip installable wheel
  • gradient_cache-1.0.0.tar.gz - Source distribution
  • Full documentation and examples

Installation

pip install gradient_cache-1.0.0-py3-none-any.whl

Acknowledgments

Built with PyTorch 2.7.1 and tested extensively on Lightning AI platform.