Final PyTorch 1.8 Release

This is the final release for PyTorch 1.8. This branch has been archived in "archive-1.8". The develop branch now supports PyTorch 2.

What's Changed

• Updated FaceID by @oguzhanbsolak in #322
• Add NAS KWS model (trained using dynamic augmentation) by @alicangok in #324
• Add Autoencoder Model by @asyatrhl in #329
• Fix pooling when the operator is None and the input is 1d by @ermanok in #343
• Scripts, Logs & Trained Model Files for MSnoise Mixed Trained NAS & v3 models by @EyubogluMerve in #332
• Final PyTorch 1.8 release - freeze requirements by @rotx-maxim in #349
• Documentation updates
• Bug fixes

Full Changelog: 2023121...2024070

Updated and new pre--trained weights; bug fixes and documentation updates

What's Changed

• Bug fixes and documentation updates
• New and updated pre-trained weights for Face detection, Feature Pyramid Detector - PASCAL VOC, Action Recognition model, Bayer2RGB model, Cats&Dogs model, KWSv3
• Add support for cropping (in_crop)
• Ubuntu 22.04 support

Bug fixes, documentation updates, weight deduplication, latency estimates

What's Changed

• Bug fixes
• Documentation updates
• Link to latest SDK
• Allow multiple output layers to have mixed output widths
• Add latency estimates for MAX78000
• Add weight deduplication for MAX78000 and MAX78002
• Update VSCode & Makefile project generators

Full Changelog: 2022071...2022122

Object detection model; various improvements

• Object detection model and example
• Bug fixes
• Improved default settings
• Improved documentation
• Updated examples
• Updated submodules

Updates to SDK, MAX78002, Documentation

• SDK updates
• MAX78002 support / bug fixes
• Improved documentation

Please follow the instructions in README.md to upgrade (ignore the zip files under “Assets”).

Improved documentation and Windows support, bug fixes

Note: “develop” is now the default branch. The virtual environment is now in the “venv” folder.

• Documentation improvements
• Bug fixes
• Windows support (WSL2 and native)
• Dilated Conv1d
• Additional models and datasets
• Updated SDK

Please follow the instructions in README.md to upgrade (ignore the zip files under “Assets”).

Improved documentation, bug fixes

• Documentation improvements
• Bug fixes
• Built-in update checker

Please follow the instructions in README.md to upgrade (ignore the zip files in the Release).

Updated SDK and Documentation, Bugfixes

•	Update documentation
•	Update SDK
•	Update models
•	Improve kernel memory allocation
•	Bug fixes

Please follow the instructions in README.md to upgrade (ignore the zip files in the Release).

PyTorch 1.8.1, updated SDK and Documentation, Bugfixes

• Update documentation
• Update SDK
• Upgrade to PyTorch 1.8.1
• ConvTranspose2d support
• Update models
• Improve bias memory allocation
• Bug fixes

Please follow the instructions in README.md to upgrade (ignore the zip files in the Release).

QAT, updated SDK, Bugfixes

The main improvement in this release is enabling Quantization-Aware Training (QAT) by default as well as the latest SDK, and changes to the generator that separate the wrapper code from the CNN code.
Additionally, it upgrades to:

• PyTorch 1.71, Python 3.8.6, CUDA 11
  There is also a number of bug fixes and documentation improvements.

Please follow the instructions in README.md to upgrade (ignore the zip files in the Release).