v0.2.6

Full Changelog: v0.2.4...v0.2.6

v0.2.4

bumped version number

v0.2.3

Added support for vector dequantization (with a vector size of 2)

v0.1.0

What's Changed

• Add learnable scales functionality. by @radi-cho in #8
• Better HuggingFace support.

New Contributors

• @radi-cho made their first contribution in #8

Full Changelog: v0.0.7...v0.1.0

v0.0.7

Added bitsandbytes conversion support.

v0.0.6

Patch release.

Full Changelog: v0.0.5...v0.0.6

v0.0.5

Added support for RTX4090

v0.0.4

1. Adding support for LLaMA-3.1 405B
2. Lightly tuned BF16 performance, though still worse than FP16, especially in 3-bit settings.
3. Uses newer vLLM version.

v0.0.3

included dependencies in setup

v0.0.2

Renamed the distribution name to flute-kernel to avoid name conflicts.