All pull requests

Pull requests

• Skip prodigy and dadapt_adamw steps until a non-zero gradient is seen

  #1820·sridhar-3009 opened Oct 13, 2026·

  ·

• Fix NaN from polynomial_schedule with fractional power under jit

  #1819·sridhar-3009 opened Oct 13, 2026·

  ·

• Initialize the first moment of scale_by_yogi to zero

  #1818·satyam-mishra-dev opened Oct 13, 2026·

  ·
  1

  1

• fix(clipping): keep zero-update reverse gradients finite

  #1815·jabrailkhalil opened Oct 12, 2026·

  ·
  1

  1

• fix:Add Self-Scaled Broyden (SSBroyden) and SSBFGS optimizers (#1658)

  #1812·Aditya-B-007 opened Oct 9, 2026·4 tasks done·

  ·

• Fix reduce_on_plateau learning rate reduction with patience=0

  #1811·judyhe19 opened Oct 8, 2026·

  ·
  1

  1

• Fix overflow in generalized Dice class weights

  #1809·Robertboy18 opened Oct 6, 2026·

  ·

• fix: RMS transforms produce NaN on float16 zero-gradient steps

  #1808·Aditya-B-007 opened Oct 6, 2026·

  ·
  2

• Reduce zero-patience schedules on plateaus rather than improvements

  #1806·sylvesterkaczmarek opened Oct 6, 2026·

  ·
  1

  1

• docs: add Simplified Chinese translation (README.zh.md)

  #1804·JasonYeYuhe opened Oct 6, 2026·

  ·

• Keep ignored ranking entries out of loss gradients

  #1803·sylvesterkaczmarek opened Oct 5, 2026·

  ·

• Mask cosine inputs before normalization to preserve finite gradients

  #1801·sylvesterkaczmarek opened Oct 3, 2026·

  ·
  1

  14

• Keep zero entries on the sphere in projection_l1_sphere

  #1799·vineethsaivs opened Oct 1, 2026·

  ·

• Stabilize NT-Xent loss and gradients at low temperatures

  #1798·Ayushdevo opened Oct 1, 2026·

  ·

• Fix NaNs in zero-radius L2 ball projections

  #1797·Ayushdevo opened Oct 1, 2026·

  ·
  2

• Apply polyak_sgd's scaling after the Polyak step

  #1796·vineethsaivs opened Sep 30, 2026·

  ·

• Keep zoom linesearch state types fixed so a jitted lbfgs update compiles once

  #1795·mahdi-shafiei opened Sep 25, 2026·

  ·

• Take the absolute difference in triplet_margin_loss

  #1794·vineethsaivs opened Sep 24, 2026·

  ·

• Stabilize integer-label softmax cross-entropy gradients under high model confidence.

  #1793·copybara-service[bot] opened Sep 24, 2026Bot·

  ·

• Stabilize integer-label cross entropy under logit shifts

  #1792·sylvesterkaczmarek opened Sep 22, 2026·

  ·
  1

• Stop adafactor's step_offset from producing a non-finite decay rate

  #1791·vineethsaivs opened Sep 21, 2026·

  ·

• Fix float16 zero-gradient NaNs in AdaDelta, Adan, and NovoGrad

  #1790·Anu27n opened Sep 18, 2026·1 of 2 tasks·

  ·
  1

  7

• Fix sam with sync_period=1 undoing the previous step in transparent mode

  #1788·Nicholas022400701 opened Sep 18, 2026·

  ·
  2

• Fix the weight decay update in momo and momo_adam to match the paper and the reference implementation

  #1787·Nicholas022400701 opened Sep 18, 2026·

  ·
  1

• Keep kl_divergence finite when target probabilities are zero

  #1786·Nicholas022400701 opened Sep 17, 2026·

  ·