September 22, 2026
- Add Qwen-Drive-1.0-4B vision tower weights.
- Fix prefetch loader stream reuse/synchronization and AugMix for non-square and single-band images.
- Fix AdaMuon conv LR scaling, AdafactorBigVision clipping, MARS gradient history, and Kron checkpoint loading.
- Fix stale eval attention bias caches, pruned model feature metadata, dilated max-pool padding, and asymmetric loss gradient mode.
- Release 1.0.30
September 10-11, 2026
- Add Qwen3-VL / Qwen3.5 / Qwen3.8 ViT classifier and encoder variants, including classifiers w/ and w/o the native spatial merger.
- Add Sapiens2 ViTs with EVA and NaFlexViT support. Both model additions include native timm Hub weights. Thanks Yonghye Kwon.
- Add iFormer with native timm Hub weights. Thanks Ryan Hou.
- Add EfficientViM with native timm Hub weights. Thanks smellslikeml.
- Add DeepSeek-V4 / V4.1 ViT classifiers and encoders with native timm Hub weights. Thanks Yonghye Kwon.
- Switch default NaFlex SigLIP position interpolation,
inference.pyinput-size selection, and improve--model-kwargsparsing. - Fix meta-device construction and device/dtype factory support for a few models; move config tests to meta for faster CI, meta context coverage.
What's Changed
- Fix ParseKwargs crash on a --model-kwargs value containing '=' or a non-literal by @winklemad in #2758
- Sapiens2 by @rwightman in #2768
- Add Sapiens2 (Meta human-centric ViT) to Eva: GQA, attn-only LayerScale, 5 model defs by @developer0hye in #2763
- Fix inference test-time preprocessing defaults by @Pdevadiga45 in #2769
- Qwen3 vit by @rwightman in #2771
- Add Qwen3.5 / Qwen3.8 (Qwen3-VL) vision encoder as qwen3_vit by @developer0hye in #2765
- Integrate iformer & efficientvim by @rwightman in #2776
- Add iFormer by @brianhou0208 in #2757
- Add EfficientViM by @smellslikeml in #2767
- Integrate deepseek vit by @rwightman in #2777
- Add DeepSeek-V4 / V4.1 vision encoder as deepseek_vit by @developer0hye in #2775
- Fix models that cannot be created with 'meta' device context. by @rwightman in #2778
- Fix #2781, CUDA stream allocation by @rwightman in #2784
- fix augmix mixing buffer shape for non-square and single-band images by @caiotheodoro in #2783
- chore: enable Dependabot weekly GitHub Actions bumps by @hf-dependantbot-rollout[bot] in #2703
- Bump the actions group with 7 updates by @dependabot[bot] in #2787
- Preserve caller gradient mode in AsymmetricLossMultiLabel by @TomasGuija in #2760
- Add Qwen-Drive-1.0-4B vision tower as a qwen3_vit pretrained tag by @developer0hye in #2774
- Copy the previous gradient in Mars instead of keeping a reference to p.grad by @Nicholas022400701 in #2786
- Fix SAME padding for dilated max pooling by @trueoneplusone in #2779
- Fix feature_info channels for pruned models (#2370) by @rwightman in #2789
- Fix RMS clipping direction in AdafactorBigVision by @Nicholas022400701 in #2790
- Fix AdaMuon LR scale for conv weights by @Nicholas022400701 in #2791
- Fix shadowed Kron setstate so loaded checkpoints get corrected_weight_decay by @Nicholas022400701 in #2792
- Clear the eval-mode attention bias cache on every train()/eval() call by @YHC66 in #2795
New Contributors
- @Pdevadiga45 made their first contribution in #2769
- @smellslikeml made their first contribution in #2767
- @caiotheodoro made their first contribution in #2783
- @hf-dependantbot-rollout[bot] made their first contribution in #2703
- @TomasGuija made their first contribution in #2760
- @Nicholas022400701 made their first contribution in #2786
- @trueoneplusone made their first contribution in #2779
- @YHC66 made their first contribution in #2795
Full Changelog: v1.0.29...v1.0.30