Issues

• Out-of-bounds codebook read in the Triton gemv kernel when a checkpoint codes value exceeds codebook_size (CWE-125)

  Vahe1994/AQLM#191 · Filter by author professor-moody opened on Jul 14, 2026

• Embedding int8: -375MB, cosine sim 0.9999

  Vahe1994/AQLM#190 · Filter by author thedorohindorohin-ctrl opened on May 23, 2026

• 30+ compression experiments on AQLM codes — all failed. One thing worked.

  Vahe1994/AQLM#189 · Filter by author thedorohindorohin-ctrl opened on May 23, 2026

• Add devstral small 2 (mistral3) support

  Vahe1994/AQLM#185 · Filter by author lexasub opened on Feb 15, 2026

• How to get vector quantized in output_dim?

  Vahe1994/AQLM#184 · Filter by author shanhx2000 opened on Oct 20, 2025

• Compatibility with Qwen models

  Vahe1994/AQLM#183 · Filter by author AIpplier opened on Oct 16, 2025
  1

• Release new version as Intel XPU supported

  Vahe1994/AQLM#182 · Filter by author jiqing-feng opened on Aug 20, 2025
  1

• AQLM for VLM

  Vahe1994/AQLM#181 · Filter by author LugerW-A opened on Jul 18, 2025
  1

• Only Use AQ part

  Vahe1994/AQLM#178 · Filter by author zzzzzzzyb opened on Jun 1, 2025
  2

• he calculation of the compression ratio

  stale
  Vahe1994/AQLM#177 · Filter by author maoshanwen opened on Apr 30, 2025
  2

• Is two-bit quantization only achievable under specific conditions?

  stale
  Vahe1994/AQLM#174 · Filter by author maoshanwen opened on Apr 19, 2025
  2

• Quantized model reverts to original size after converted to HF format

  stale
  Vahe1994/AQLM#173 · Filter by author QianChef opened on Apr 19, 2025
  1