GV

GeezerVison

The World Wide Web, restored to 800 pixels of glory.

Home|About|Gaming cabinet|Archives|How it works|Starter sites

On this site

  • Sign in
  • ActionsCI/CD and automation
  • CodespacesCloud development environments
  • All Features
  • Documentation
  • Training
  • Certifications
  • Build with Copilot
  • Customize Copilot
  • Integrations
  • Trending repos
  • GitHub Sponsors
  • Developer articles
  • Enterprises

This view

  • Original page
  • New address

GeezerVison is showing github.com — Pull requests · bitsandbytes-foundation/bitsandbytes

Pull requests

Open
66 (66)
Closed
718 (718)
More items
  • Comfortable display density
  • Compact display density

  • Move cached int8 linear weights when moving parent modules

    #2129·tgghzyh opened Nov 12, 2026·
    ·

  • Avoid copying the full scale table in 4-bit embedding lookups

    #2128·tgghzyh opened Nov 11, 2026·
    ·

  • perf: avoid repeated INT8 cast-warning records

    #2127·yfukuda1-oqc opened Nov 11, 2026·
    ·

  • docs: fix the LICENSE badge link

    #2125·cloudgpuapp opened Nov 10, 2026·
    ·

  • fix(gaudi): default the plugin version when metadata has no Version

    #2124·devinhuang2001 opened Nov 9, 2026·
    ·

  • test(mps): verify 4-bit dequantization values and Hub dispatch

    #2123·egeozkoc opened Nov 8, 2026Contributor·
    ·
    1

  • chore(deps): bump the actions group across 1 directory with 5 updates

    #2120·dependabot[bot] opened Nov 2, 2026Bot·
    ·
    1

  • Fix CPU quantize_blockwise to pick the nearest code

    #2119·damienen opened Nov 2, 2026·
    ·
    1

  • Do not freeze zero-norm parameters in LAMB/LARS trust-ratio clipping

    #2117·Andreas-Hum opened Nov 1, 2026·
    ·
    1

  • Fix INT8 weight-gradient quantization and CUDA operand layout

    #2115·devinhuang2001 opened Oct 30, 2026·
    ·
    1

  • Mock torch.version.hip in CUDA override tests

    #2114·kudomcho opened Oct 29, 2026·2 of 3 tasks·
    ·
    1

  • Fix fp32 4-bit GEMV test tolerance for ROCm/gfx1200

    #2113·kudomcho opened Oct 29, 2026·3 of 4 tasks·
    ·
    1

  • Fix k-bit bias gradients for non-2D inputs

    #2111·anh-nguyenn opened Oct 26, 2026·
    ·
    1

  • Fix MatMul8bitFp backward for non-2D inputs

    #2109·anh-nguyenn opened Oct 26, 2026·
    ·
    1

  • Validate packed size against quant_state.shape in dequantize_4bit

    #2107·tritsystem opened Oct 25, 2026·
    ·
    1
    1

  • Fix quantize_4bit under-allocating the packed buffer for wide quant_storage

    #2105·tritsystem opened Oct 25, 2026·
    ·

  • Fix CUDA INT8 quantization of non-contiguous inputs

    #2104·zupengwang opened Oct 23, 2026·
    ·

  • Make the Windows ROCm load path work and fix has_imma on ROCm

    #2102·IxMxAMAR opened Oct 18, 2026·
    ·

  • Cache the ROCm warp size per device instead of for device 0

    #2101·IxMxAMAR opened Oct 18, 2026·
    ·

  • Fix XPU quantize_4bit returning padded tensors that violate the op contract

    #2099·jiqing-feng opened Oct 15, 2026Contributor·
    ·
    · v0.50.3
    3

  • Optimize NF4 single-token GEMV on Blackwell

    #2098·PiotrPotocki opened Oct 8, 2026·
    ·

  • Support skip_zeros in the XPU Triton optimizer kernels

    #2097·jiqing-feng opened Oct 8, 2026Contributor·
    ·
    · v0.50.3
    5

  • fix: wrong absmax indexing in AVX512 dequantizeBlockwise4bitCpu

    #2096·NanoCode012 opened Oct 7, 2026·
    ·

  • MPS: only use the Hub GEMV kernel when K is a multiple of blocksize

    #2094·egeozkoc opened Sep 30, 2026Contributor·
    ·
    1

  • test: align post-reload max_error_count with pre-reload check in test_optimizer32bit

    #2093·CS-liujf opened Sep 23, 2026·
    ·
You are visitor 01402503 — GeezerVison does not store the pages you look up.
Best viewed with Netscape Navigator 4.7 or Internet Explorer 6 at 800×600. No cookies required.