-
Notifications
You must be signed in to change notification settings - Fork 370
Pull requests: google-deepmind/optax
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Fix float16 zero-gradient NaNs in AdaDelta, Adan, and NovoGrad
#1790
opened Sep 18, 2026 by
Anu27n
Loading…
1 of 2 tasks
Fix sam with sync_period=1 undoing the previous step in transparent mode
#1788
opened Sep 18, 2026 by
Nicholas022400701
Loading…
Fix the weight decay update in momo and momo_adam to match the paper and the reference implementation
#1787
opened Sep 18, 2026 by
Nicholas022400701
Loading…
Keep kl_divergence finite when target probabilities are zero
#1786
opened Sep 17, 2026 by
Nicholas022400701
Loading…
Ensure CTC loss is nonnegative against float32 rounding errors (fixes #1771)
#1783
opened Sep 15, 2026 by
jdymitarai
Loading…
Fix onecycle and warmup cosine schedule docstrings to match the code
#1781
opened Sep 12, 2026 by
rishabhsai
Loading…
Preserve plateau scale dtype with array minimum scales
#1774
opened Sep 10, 2026 by
sylvesterkaczmarek
Loading…
Fix Pyrefly missing-attribute and unsupported-operation ignores on ArrayLike values by normalizing them to jax.Array using jnp.asarray.
#1773
opened Sep 8, 2026 by
copybara-service
Bot
Loading…
Fix negative CTC loss for confident alignments in float32
#1772
opened Sep 7, 2026 by
rajasekharporeddy
Collaborator
Loading…
fix(transform): prevent NaN in Adam-family optimizers on float16 zero-gradient steps
#1769
opened Sep 3, 2026 by
somuai
Loading…
fix(transform): prevent NaN in scale_by_adamax on float16 zero-gradient steps
#1768
opened Sep 3, 2026 by
somuai
Loading…
Fix
contrib.reduce_on_plateau never reducing the learning rate on negative values
#1767
opened Sep 2, 2026 by
Kayvan-Zahiri
Loading…
Fix a NaN in
contrib.dadapt_adamw on zero denominators and zero learning rates.
#1766
opened Aug 30, 2026 by
copybara-service
Bot
Loading…
Guard
contrib.prodigy against zero denominators and zero learning rates.
#1765
opened Aug 30, 2026 by
copybara-service
Bot
Loading…
fix(losses): avoid NaN gradient in weighted_logsoftmax at -inf logits
#1762
opened Aug 25, 2026 by
jaideeppyne
Loading…
Fix all-NaN output for complex inputs with 'schatten' preconditioning in Muon
#1761
opened Aug 24, 2026 by
shoemoney
Loading…
docs: fix typos and update docstring formatting across multiple modules
#1759
opened Aug 19, 2026 by
rajasekharporeddy
Collaborator
Loading…
Fix NaN in scale_by_adam on float16 zero-gradient steps
#1755
opened Aug 16, 2026 by
kartik1pandey
Loading…
Add SPSA (Simultaneous Perturbation Stochastic Approximation) optimizer to contrib
#1753
opened Aug 16, 2026 by
jaideeppyne
Loading…
Fix clip_by_global_norm producing NaN when the global norm and max_norm are both zero
#1752
opened Aug 15, 2026 by
shashvat-singham
Loading…
Previous Next
ProTip!
Type g i on any issue or pull request to go back to the issue listing page.