Add k-bit blockwise quantization (K=2-5) with warp-level CUDA kernels#1858

Open

TimDettmers wants to merge 11 commits intomainfrom

feature/kbit-quantization

Commits on Feb 14, 2026

Add k-bit quantization kernels (K=2-5, blocksize=32) -- WIP

TimDettmers
and
claude
committed
Fix RDC device linking: move kernels to ops.cu, all 157 tests pass

TimDettmers
and
claude
committed
Complete k-bit quantization: Stages 6-8, Python API, 218 tests pass

TimDettmers
and
claude
committed
Remove implementation progress report

TimDettmers
and
claude
committed
Add vectorized dequant kernel and E4M4 uint8 absmax support

TimDettmers
and
claude
committed
Remove scalar dequant kernel, fp32 absmax, and Stage 1-3 scaffolding

TimDettmers
and
claude
committed
Template dequant kernel on output type, add bf16/fp32 native output

TimDettmers
and
claude
committed
Fix lint and formatting issues from CI pre-commit checks

TimDettmers
and
claude
committed
Fix analytical error bound for K=5 with E4M4 absmax

TimDettmers
and
claude
committed

Commits on Feb 22, 2026