Apache SINGA-5.1.0 Release Notes
SINGA is a distributed deep learning library.
This release includes following changes:
Add the implementations for Low-Rank Adaptation of LLMs (LORA).
Add the implementations for Parameter-Efficient Fine-Tuning (PEFT).
- Add the implementations of distributed models for SINGA PEFT.
- Add the implementations of tuners for SINGA PEFT.
- Add the implementations of the autograd functions for SINGA PEFT.
- Add the multiprocess implementations for SINGA PEFT.
Update the healthcare model zoo.
- Restructure the folders for the healthcare model zoo.
- Update the training files and model files for the healthcare model zoo.
- Update the documentations for the healthcare model zoo.
Add more healthcare examples.
- Add the implementations of the candidiasis disease application.
- Add the implementations of the cardiovascular disease application.
Add the implementations for more attention mechanisms.
Update the encoder layers for the transformer model.
Fix typos in Python module docstrings.
