Release Notes#

This page documents the updates for each version of Kaiwu-PyTorch-Plugin.

Version 0.1.0 (2025-12-24)#

New Features

  • Dependency license check

  • PyPI release pipeline

Improvements

  • Documentation and instructions updated

  • Adjusted BM interface and matrix conversion logic

Version 0.0.2 (2025-11-06)#

New Features

  • Added documentation framework with Sphinx support

  • Added DBN (Deep Belief Network) module supporting multi-layer RBM stacking

  • Added Q-VAE example supporting quantum variational autoencoder training

Improvements

  • Restored and optimized RBM module implementation

  • Updated example code README

  • Improved code comments for better readability

Documentation

  • Updated Chinese and English README files

  • Added more detailed installation instructions

  • Supplemented example code descriptions

Bug Fixes

  • Fixed known issues in the DBN module

  • Fixed code issues detected by pylint

Version 0.0.1 (2025-10-15)#

Initial Release

The first official release of Kaiwu-PyTorch-Plugin, providing the following core features:

Core Modules

  • RestrictedBoltzmannMachine: Implementation of Restricted Boltzmann Machine

  • BoltzmannMachine: Implementation of fully-connected Boltzmann Machine

  • AbstractBoltzmannMachine: Abstract base class supporting custom extensions

Key Features

  • Supports native PyTorch interface, compatible with standard optimizers (SGD, Adam, etc.)

  • Supports Kaiwu SDK samplers, including simulated annealing and quantum sampling

  • Supports GPU-accelerated training

  • Provides complete example code

Examples

  • rbm_digits: Handwritten digit recognition example

  • qvae_mnist: Q-VAE MNIST generation example

  • bm_generation: Boltzmann Machine data generation example

Requirements

  • Python 3.10

  • PyTorch 2.7.0

  • Kaiwu SDK v1.2.0+

On this page