Skip to content

Release Notes for Split_LoRA v1.0.0

Choose a tag to compare

View all tags
@banbo1314 banbo1314 released this 16 Jul 11:34
· 48 commits to main since this release
v1.0.0
e907a34
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

[1.0.0] 2024-07-15

We are thrilled to announce the release of version 1.0.0 of the SplitFed-LoRA project! This release introduces numerous new features, performance improvements, and bug fixes, making the training of distributed machine learning and low-rank adaptive models more efficient and user-friendly.

New Features

  • Distributed Training Support: This release introduces a distributed splitfed learning framework that supports training across multiple devices and nodes, significantly enhancing model training speed and performance.
  • LoRA Integration: Incorporates low-rank adaptive methods, drastically reducing model parameter count, and making models more lightweight and suitable for resource-constrained environments.
  • Enhanced Security: Adds new data encryption and privacy protection features to ensure data security and privacy during the training process.
  • Auto-Tuning: The new automatic tuning feature dynamically adjusts training parameters based on data distribution and model performance, enhancing model accuracy.

Performance Improvements

  • Memory Optimization: Significantly reduces memory usage and enhances training efficiency on large-scale datasets by improving data loading and processing mechanisms.
  • Speed Boost: Optimizes the model training pipeline, increasing training speed and reducing overall training time.
  • Fault Tolerance Mechanism: Enhances fault tolerance with automatic task reallocation in case of node failures, ensuring robustness during the training process.

Bug Fixes

  • Fixed issues related to data loading failures under certain circumstances.
  • Corrected errors during model saving to ensure model persistence throughout the training process.
  • Addressed synchronization issues that may occur during distributed training.

Contribution

We welcome contributions and feedback from the community. If you have any questions or suggestions, please submit them via GitHub Issues or contribute directly to our repository.
Thank you once again for your support and participation!

Assets 4
Loading