bitsandbytes Files

This release adds support for additional hardware platforms and improves performance on RTX Spark / DGX Spark.

Highlights

• Support added for NVIDIA RTX Spark products running Windows on ARM64 (#2038, [#2037], [#2032])
• Improved 4bit GEMM dispatch heuristics for NVIDIA GB10 (#2030)
• Support for AMD CDNA5 hardware (e.g. MI455X)

What's Changed

• Add ROCm CDNA5 (gfx1250) support by @zjin-lcf in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2023
• chore(deps): bump the actions group with 3 updates by @dependabot[bot] in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2022
• Update README by @matthewdouglas in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2024
• Fix truncated FP4 constant in Triton dequantize_fp4_tree by @Anai-Guo in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2026
• docs: add installation instructions for NVIDIA Spark (ARM64) devices by @mfuntowicz in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2037
• fix: missing compilation flag on recent msvc version by @mfuntowicz in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2032
• build: Add Windows ARM64 build with CUDA 13.4 for RTX Spark by @matthewdouglas in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2038
• Add sm121 (GB10) tier to the 4-bit GEMM dispatch heuristic by @yashb98 in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2030

New Contributors

• @zjin-lcf made their first contribution in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2023
• @Anai-Guo made their first contribution in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2026
• @yashb98 made their first contribution in https://github.com/bitsandbytes-foundation/bitsandbytes/pull/2030

Full Changelog: https://github.com/bitsandbytes-foundation/bitsandbytes/compare/0.50.0...0.50.1