This repository was archived by the owner on Oct 3, 2026. It is now read-only.

Repository navigation

All pull requests

Labels (9)Milestones (0)

Fix an out-of-bounds read in the fused window kernels, generalise to non-square shapes, port to ROCm
#389Francesco Brigante (francescobrigante) opened Aug 29, 2026
1
Fix typo in README.md
#381Shubhranshi Agarwal (shubhranshii) opened Mar 8, 2025
swin_transformer_v2.py RuntimeError Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument max in method wrapper_CUDA_clamp_Tensor)
#376Woongjun Choi (woongjoonchoi) opened Jan 15, 2025
Fix typo in README.md
#351polar-bears (JihwanEom) opened Apr 12, 2024
Using torch.bfloat16 to prevent overflow instead of default fp16 in AMP
#345Rajeev Goel (rajeevgl01) opened Jan 12, 2024
fix swin_transformer ops computation
#339shlrao (russ-spec) opened Nov 8, 2023
minor fix, remove t_mul passing to CosineLRScheduler
#315Zengwei Yao (yaozengwei) opened May 23, 2023
Update README.md for description of FasterTransformer
#299Jackch-NV opened Dec 27, 2022Contributor
Wip configs
#295Dipanjyoti Paul (dipanjyoti) opened Dec 7, 2022
1
add swin2sr (ECCVW 2022) cross reference
#276Marcos V. Conde (mv-lab) opened Oct 2, 2022
supported dynamic batch for torchscript trace
#257dml (MenglingD) opened Aug 29, 2022
1
Add link to third-part Swin V2 & V2 TensorFlow implementation
#250Shkarupa Alex (shkarupa-alex) opened Aug 12, 2022
Fix torch.clamp issue #237
#238Hyunsoo Luke HA (CryptoSalamander) opened Jul 14, 2022
3
Update README.md - Fixed Typo
#211Adam Hackbarth (digitalvirtuoso) opened May 20, 2022
Simplifying some code
#188Amir Masoud Nourollah (Nourollah) opened Apr 3, 2022
1
Change load_pretrained function to accept models from MoBY
#184Giles Billenness (Giles-Billenness) opened Mar 16, 2022
2
skip .view() func in WindowAttention
#119Yonghye Kwon (developer0hye) opened Sep 3, 2021

Pull requests