Search results

VariationalDropout modules do not support CUDA device

Status: Open.

#13 In locuslab/torchdeq;

· lcdavis13 opened on Apr 5, 2026

stop_mode='rel' seems to be ignored (solver iterates until 1e-5 despite 5% tolerance)

Status: Open.

#12 In locuslab/torchdeq;

· QuietNoviceCoder opened on Nov 17, 2025

torch.compile failed due to modified backward pass

Status: Open.

#11 In locuslab/torchdeq;

· sablin39 opened on Nov 14, 2025

jac_loss has no gradient for training.

Status: Open.

#10 In locuslab/torchdeq;

· QuietNoviceCoder opened on Oct 29, 2025

[New Feature] torchdeq implementation for *Reversible Deep Equilibrium Models*

Status: Open.

#9 In locuslab/torchdeq;

· sablin39 opened on Sep 27, 2025

Returned fixed point solution sometimes has requires_grad=False

Status: Open.

#8 In locuslab/torchdeq;

· Boshi-Wang opened on Jun 27, 2024

Fixed-point is not returned when indexing is set

Status: Open.

#7 In locuslab/torchdeq;

· BurgerAndreas opened on May 30, 2024

Unexpected behaviour of indexing?

Status: Open.

#6 In locuslab/torchdeq;

· BurgerAndreas opened on May 25, 2024

Custom autograd fails with torchdeq in eval mode

Status: Open.

#4 In locuslab/torchdeq;

· BurgerAndreas opened on Mar 31, 2024

Implementation of torchdeq.utils.mixed_init different from original paper

Status: Open.

#3 In locuslab/torchdeq;

· jklim01 opened on Mar 15, 2024