tch

Changed

• PyTorch v2.7 support

Changed

• PyTorch v2.6 support

Changed

• PyTorch v2.5.1 support

Changed

• PyTorch v2.5 support

Changed

• PyTorch v2.4 support

Changed

• PyTorch v2.3 support

Changed

• PyTorch v2.2 support

Changed

• PyTorch v2.1 support 803.
• Add a pyo3-tch crate for interacting with Python via PyO3 730.
• Expose the cuda fuser enabled flag, 728.
• Improved the safetensor error wrapping, 720.

Added

• Support static linking in the build script, 712.
• Make the libtorch download opt-in rather than a default behavior. The libtorch library download can still be triggered by enabling the download-libtorch feature, 707.
• Rename the of_... conversion functions to from_... so as to be closer to the Rust best practices, 706. This is a breaking change and will require modifying calls such as of_slice to be from_slice instead.
• Expose some functions so that Python extensions that operates on PyTorch tensors can be written with tch, 704.
• Rework the torch-sys build script making it easier to leverage a Python PyTorch install as a source for libtorch, 703.

Changed

• EfficientNet models have been reworked, pre-trained models used safetensors weight by default, 679.
• None can be used for nullable scalar types, 680.
• Automated conversion of list arguments: all the generated functions that take as input a slice of int or float can now be used directly with int values or fixed length arrays 682.
• Replace the From<Tensor> traits with some TryFrom versions, 683. This is a breaking change, note that also the old version would flatten the tensor if needed to reduce the number of dimensions, this has to be done explicitely with the new version.

Added

• Adapt to C++ PyTorch library (libtorch) version v2.0.0.

Changed

• Update the half dependency to version 2, 646.

Added

• Add some helper functions in a utils module to check the available devices and versions.

Added

• Add the eps and amsgrad options to Adam, 600.

Changed

• Fix loading of VarStore when using Mps devices, 623.
• Use ureq instead of curl to reduce compile times, 620.
• Fix the handling of dicts in TorchScript, 597.

Changed

• Default vs.load to use the Python weight format when the file extension is .pt or .bin.

Added

• Expose functions for setting manual seeds for CUDA devices, #500.
• Expose functions for triggering manual sync, #500.
• Add some functions to load Python weight files.

Changed

• Extending the Kaiming initialization, #573.

Changed

• Adapt to C++ PyTorch library (libtorch) version v1.13.0.

Changed

• Adapt to C++ PyTorch library (libtorch) version v1.12.0.

Changed

• Adapt to C++ PyTorch library (libtorch) version v1.11.0.

Changed

• Adapt to C++ PyTorch library (libtorch) version v1.10.0.

Changed

• Adapt to C++ PyTorch library (libtorch) version v1.9.0.

Changed

• Adapt to C++ PyTorch library (libtorch) version v1.8.1.