All pull requests

Pull requests

• [ROCm] Move the ROCm CI to ROCm 7.14 and 10.0

  #173·aryaman-gupta opened Sep 30, 2026·

  ·

• [ROCm][CI] Restore ROCm GPU CI: switch to the MI350 runner label

  #168·aryaman-gupta opened Aug 19, 2026·

  ·

• [ROCm] Make the Python NFP8 dtype selection architecture-aware

  #167·aryaman-gupta opened Aug 19, 2026·

  ·
  2

• [ROCm] Fix packed-bag pooling truncation in nbit inference forward

  #166·aryaman-gupta opened Aug 19, 2026·

  ·
  5

  

• fix ssd cache assoc wave32

  #162·aryaman-gupta opened Jun 17, 2026·

  ·
  1

  

• Optimizing writes in nobag inference kernel

  #157·aryaman-gupta opened May 15, 2026·

  ·
  4

• Generalizing the HIP-optimized warp_per_row kernel

  #146·aryaman-gupta opened Mar 19, 2026·

  ·

• Split Kernel Optimization for group_index_select_or_add_2d_kernel

  enhancement
  #142·aryaman-gupta opened Feb 2, 2026·

  ·
  20

  

• Optimize group_index_select_or_add_2d_kernel by adding a separate codepath for small embedding dimensions

  #135·aryaman-gupta opened Dec 16, 2025·

  ·
  

• Fixes bug in one specialized HIP instantiation of the warp-per-row kernel

  #134·aryaman-gupta opened Dec 5, 2025·

  ·
  1