Pull Requests huggingface/diffusers

Add warning when using CUDA random generator (#13298)

#13309 opened 2026-03-22 18:03 by r266-tech

fix: disable non-blocking tensor copies to MPS during model loading

#13308 opened 2026-03-22 09:21 by agarwalprakhar2511

fix(dreambooth): batch size mismatch with --with_prior_preservation in flux2 scripts

#13307 opened 2026-03-22 09:13 by agarwalprakhar2511

fix: skip pin_memory and non_blocking transfer for tensor subclasses in group offloading

#13305 opened 2026-03-22 00:34 by s-zx

fix: support torchao >= 0.16.0 by importing renamed CamelCase Config classes

#13304 opened 2026-03-22 00:04 by s-zx

fix: cast input to weight dtype in PatchEmbed and FluxTransformer2DModel to prevent dtype mismatch

#13303 opened 2026-03-21 23:55 by s-zx

fix(attention): download hub kernel in AttentionModuleMixin.set_attention_backend

#13302 opened 2026-03-21 23:33 by s-zx

Fix missing latents_bn_std dtype cast in VAE normalization

#13299 opened 2026-03-21 09:54 by adi776borate

[Neuron] Add AWS Neuron (Trainium/Inferentia) as an officially supported device

#13289 opened 2026-03-19 16:01 by JingyaHuang

Add low precision attention API from torchao to TorchAoConfig

#13285 opened 2026-03-19 00:41 by howardzhang-cv

Update attention_backends.md to update FA3 minimum support to Ampere

#13283 opened 2026-03-18 10:14 by sayakpaul

[core] fix group offloading when using torchao roadmap

#13276 opened 2026-03-17 05:16 by sayakpaul

fix: wrap validation inference with torch.no_grad() in dreambooth examples

#13273 opened 2026-03-16 14:13 by gambletan

fix: compute image_seq_len from spatial dims in Lumina2 pipeline

#13272 opened 2026-03-16 14:10 by gambletan

Add DiffSynth blockwise ControlNet support to QwenImageControlNetModel

#13268 opened 2026-03-15 05:17 by aryabyte21

Fix missing gradient_checkpointing init in ZImageControlNetModel

#13267 opened 2026-03-15 03:43 by yunjoochoi

add tensor and numpy array support for image input in Flux2KleinPipeline

#13265 opened 2026-03-14 13:12 by abhixb

[not for land yet]: improve cuda graph support for Qwen-Image

#13263 opened 2026-03-12 20:15 by vkuzo

[CI] Refactor LTX Transformer Tests

#13254 opened 2026-03-11 10:25 by DN6

[CI] Refactor Z Image Transformer Tests

#13253 opened 2026-03-11 06:04 by DN6

Fix double-shifted sigma_min/sigma_max in FlowMatchEulerDiscreteScheduler

#13246 opened 2026-03-10 10:45 by kvr06-ai

Fix IndexError in HunyuanVideo I2V pipeline

#13244 opened 2026-03-10 10:35 by kaixuanliu

[tests] migrate test_hooks.py to pytest

#13242 opened 2026-03-10 06:33 by sayakpaul

fix: deduplicate timesteps in DPMSolverMultistep for squaredcos_cap_v2 + karras/lu sigmas

#13240 opened 2026-03-10 02:02 by Chase-Xuu

Add support for FLUX.2 Klein 9B and 4B in single_file_utils

#13237 opened 2026-03-09 22:34 by eliemichel

Fix error message in gguf_quantizer.py

#13236 opened 2026-03-09 21:49 by eliemichel

[tests] refactor caching tests.

#13235 opened 2026-03-09 13:57 by sayakpaul

[tests] migrate group offloading tests to pytest

#13234 opened 2026-03-09 13:21 by sayakpaul

Add RAE Diffusion Transformer inference/preliminary training pipelines

#13231 opened 2026-03-09 04:49 by plugyawn

Refactor `AttentionProcessorSkipHook` to Support Custom STG Logic

#13220 opened 2026-03-07 00:36 by dg845