All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Adds a
fake_cuda
parameter tofake_mode()
that allows constructing fake CUDA tensors even if CUDA is not available.
- Moves to PyTorch v1.12
- Adds support for Python 3.10
- Addresses a minor bug in Fake tensor caused by the API changes in PyTorch
- Initial release with Fake Tensor and Deferred Module Initialization