LoreonLabsPlatform
DocsHome
  • Overview

Intelligence

  • Markets
  • Builders
  • Research
  • Ecosystems
  • Launchpads
  • Search
Ecosystems

Other

svdquant-kernels

Cross-architecture CUDA kernels for SVDQuant (W4A4 with low-rank correction)

OtherEmerging
GitHubWebsite
Stars
—
Forks
—
Contributors
2
Last push
28d ago

Recent commits

Latest commits.

  • docs(blog): follow README — ncu Tensor Pipe % at one shape, drop cross-arch MFU
    eeed047ultism28d ago
  • docs(perf): replace cross-arch MFU with ncu Tensor Pipe % for nunchaku compare
    f491c1eultism29d ago
  • feat(gemm_w4a4): SVDQuantW4A4Linear nn.Module + NVFP4 nunchaku weight bridge
    5cb5a97ultism29d ago
  • docs(gotchas): tcgen05.mma kind switch — pair-pipeline rule is about overlap, not ordering
    7885f31ultism1mo ago
  • feat(gemm_w4a4 v2): add lora_schedule knob (interleave|end_grouped)
    a31d05aultism
1mo ago
  • docs(blog): correct §6.3 ordering claim; redraw β figure
    1cc206eultism1mo ago
  • docs(pages): scaffold GitHub Pages from /docs for the blog
    ba83ed6ultism1mo ago
  • docs(blog): retitle to spell out the kernel shape; add repo pointer
    ba57512ultism1mo ago

    • Top contributors

    Builders behind this project.

    ultism
    116 commits
    claude
    2 commits