All pull requests

Pull requests

• Fix EOS confidence suppression at nonzero temperature

  #139·primorLee opened Aug 22, 2026·

  ·
  1

  1

• Optimize iLLaDA end-of-thinking logit boost

  #138·mercurystraw opened Aug 21, 2026·

  ·
  2

• Uq bench

  #127·Linwei94 opened Mar 6, 2026·

  ·

• experiments

  #126·Linwei94 opened Mar 6, 2026·

  ·

• Add VRAM efficient Gumbel sampling

  #124·sean-lamont opened Feb 6, 2026·

  ·
  1

• fix generation batching and single process accelerator bugs

  #110·adistomar opened Sep 19, 2025·

  ·

• Fix bugs for eval_llada.py

  #91·Kamichanw opened Jul 5, 2025·

  ·

• generate: skip unnecessary unmask steps and warn user about it

  #76·Seas0 opened Jun 5, 2025·

  ·

• 4-bit llada

  #54·1773226512 opened Mar 23, 2025·

  ·
  2

• Supporting MPS based inference

  #16·AmanPriyanshu opened Feb 27, 2025·

  ·
  1