v0.6.0

What's Changed

• Enable torchcache to also decorate functions (+kwargs support) by @meakbiyik in #5

Full Changelog: v0.5.2...v0.6.0

v0.5.2

Fixed

• A logging issue that occasionally leads to an IndexError 98ca9eb

Full Changelog: v0.5.1...v0.5.2

v0.5.1

Added

• Automated version bumps with the Github release tag

Full Changelog: v0.5.0...v0.5.1

v0.5.0

Added

• disabled argument to programmatically enable/disable torchcache, or selectively enable for some instances of a module.

Full Changelog: v0.4.1...v0.5.0

v0.4.1

Changed

• Small updates to the documentation
• Minor refactoring

Full Changelog: v0.4.0...v0.4.1

v0.4.0

Added

• Use the mmap of Python to allow using mmapped loading for all versions of PyTorch, even with compression.

Changed

• The parameter documentation is moved from the underlying class to the exposed decorator, so that the signature hints can function properly.

Fixed

• API documentation not visible.

Full Changelog: v0.3.2...v0.4.0

v0.3.2

Fixed

• Performance regression due to weights_only in torch.load
• Allowed changing embedding shape on-the-fly

Full Changelog: v0.3.1...v0.3.2

v0.3.1

Empty release to fix publish workflow, see v0.3.0 release for the updates: https://github.com/meakbiyik/torchcache/releases/tag/v0.3.0

v0.3.0

Added

• use mmap in torch.load if available (currently in nightly: pytorch/pytorch#102549)
• finer controls for module cache: ability to set it manually it to prevent cache invalidation by mistake

Changed

• ztsd compression instead of brotli for faster decompression

Fixed

• Invalidate cache if some crucial torchcache requirements change (subsample count and compression)

Full Changelog: v0.2.0...v0.3.0

v0.2.0

Full Changelog: v0.1.0...v0.2.0