0.45.1
Overview
This is a patch release containing compatibility fixes.
Highlights
- Compatibility for
triton>=3.2.0
- Moved package configuration to
pyproject.toml
- Build system: initial support for NVIDIA Blackwell B100 GPUs, RTX 50 Blackwell series GPUs and Jetson Thor Blackwell.
- Note: Binaries built for these platforms are not included in this release. They will be included in future releases upon the availability of the upcoming CUDA Toolkit 12.7 and 12.8.
- Packaging: wheels will no longer include unit tests. (#1478)
- Sets the minimum PyTorch version to 2.0.0.
What's Changed
- Add installation doc for bnb on Ascend NPU by @ji-huazhong in #1442
- (chore) Remove unused dotfiles by @matthewdouglas in #1445
- Remove triton.ops, copy necessary bits here by @bertmaher in #1413
- chore: migrate config files to
pyproject.toml
by @SauravMaheshkar in #1373 - cleanup: remove unused kernels/C++ code by @matthewdouglas in #1458
- (Deps) Require torch 2.x and minor updates by @matthewdouglas in #1459
- FSDP-QLoRA doc updates for TRL integration by @blbadger in #1471
- Initial support blackwell by @johnnynunez in #1481
- (build) include Ada/Hopper targets in cu118 build by @matthewdouglas in #1487
- Exclude tests from distribution by @akx in #1486
New Contributors
- @bertmaher made their first contribution in #1413
- @SauravMaheshkar made their first contribution in #1373
- @blbadger made their first contribution in #1471
- @johnnynunez made their first contribution in #1481
Full Changelog: 0.45.0...0.45.1