Search results

Silent corrupted LoRA weight updates with Opacus 1.5.4 + PEFT 0.18.x: training appears normal but models are unusable

Status: Open.

#820 In meta-pytorch/opacus;

· MN-Noor opened on May 19, 2026

Potential mismatch between FEASGM-style normalization and SGM-based privacy accounting

Status: Open.

#819 In meta-pytorch/opacus;

· Eexac opened on May 18, 2026

WeightedRandomSampler causes silent epsilon miscalculation)

Status: Open.

#813 In meta-pytorch/opacus;

· MN-Noor opened on Mar 29, 2026

Inconsistent privacy budget

Status: Open.

#804 In meta-pytorch/opacus;

· Drxdx opened on Jan 13, 2026

How to calculate the overall privacy budget generated during federated differential privacy model training using Opacus?

Status: Open.

#803 In meta-pytorch/opacus;

· will-last opened on Dec 3, 2025

Support LoRA adapters with Tensor Parallelism (TP)

Status: Open.

#789 In meta-pytorch/opacus;

· xzhouAxon opened on Oct 3, 2025

DPMultiheadAttention Doesn't check attention mask correctly when batch_first is set to True.

Status: Open.

#774 In meta-pytorch/opacus;

· Ghazi-Echo opened on Jul 1, 2025

Support for Shifted Window Attention in Opacus

enhancementNew feature or requestNew feature or request

Status: Open.

#752 In meta-pytorch/opacus;

· xoxo121 opened on Apr 22, 2025

RuntimeError: Tensor on device cuda:0 is not on the expected device meta!

Status: Open.

#737 In meta-pytorch/opacus;

· LucasMagnana opened on Mar 3, 2025

Microbatching Support

enhancementNew feature or requestNew feature or request

Status: Open.

#655 In meta-pytorch/opacus;

· shs037 opened on Jul 13, 2024

BatchMemoryManager for training private GANs

enhancementNew feature or requestNew feature or request

Status: Open.

#647 In meta-pytorch/opacus;

· Caarinaaa opened on Apr 26, 2024

ValueError: Per sample gradient is not initialized. Not updated in backward pass?

Status: Open.

#645 In meta-pytorch/opacus;

· Lhsakfheudsj35407 opened on Apr 12, 2024