Issues

• [Bug] Weighted classification loss uses a per-GPU denominator, so it depends on how images are split across GPUs

  SHI-Labs/OneFormer#145 · Filter by author gss10282025 opened on Sep 26, 2026

• Verify evals on Papers with Code

  SHI-Labs/OneFormer#144 · Filter by author NielsRogge opened on Sep 14, 2026

• Access the embeddings of detected objects

  SHI-Labs/OneFormer#143 · Filter by author AelaLS opened on May 26, 2026

• 📋 Documentation Enhancement Suggestion

  SHI-Labs/OneFormer#142 · Filter by author croviatrust opened on Mar 25, 2026

• Why does the model's performance decrease when using a pre-trained model and then training on the Mapillary Vistas dataset?

  SHI-Labs/OneFormer#141 · Filter by author 1183498834 opened on Aug 22, 2025
  1

• Evaluation Results Are Very Low Only in --eval-only Mode (AP=0.0), But Normal During Training

  SHI-Labs/OneFormer#140 · Filter by author GuanyuChen8 opened on Jun 20, 2025

• Setup CUDA Kernel for MSDeformAttn Fail

  SHI-Labs/OneFormer#139 · Filter by author dsinapova opened on Jun 16, 2025

• Setup and train OneFormer in Google Colab problems

  SHI-Labs/OneFormer#138 · Filter by author dsinapova opened on Jun 15, 2025

• A question about query-text contrastive loss

  SHI-Labs/OneFormer#137 · Filter by author zdk258 opened on Apr 2, 2025

• On a customized dataset, the AP is low, like under 10%, but the mIoU is high, like above 60%.

  SHI-Labs/OneFormer#136 · Filter by author hehongjie opened on Mar 31, 2025

• Support Multi-node ?

  SHI-Labs/OneFormer#135 · Filter by author ahatamiz opened on Mar 28, 2025
  2

• Incorrect color conversion in VisualizationDemo

  SHI-Labs/OneFormer#134 · Filter by author weitanism opened on Mar 10, 2025

• Running DEMO on multiple GPUs

  SHI-Labs/OneFormer#133 · Filter by author maffan-96 opened on Feb 25, 2025
  2

• Pip version in installation instructions to new

  SHI-Labs/OneFormer#132 · Filter by author koda1026 opened on Feb 15, 2025
  1

• FineTune on Custom instance coco dataset

  SHI-Labs/OneFormer#131 · Filter by author mathisTH opened on Feb 8, 2025

• Pixel decoder internal structure

  SHI-Labs/OneFormer#130 · Filter by author MicheleCazzola opened on Nov 30, 2024

• Classes indices

  SHI-Labs/OneFormer#128 · Filter by author MalekSamet opened on Oct 29, 2024

• Question about performance of smaller backbones

  SHI-Labs/OneFormer#127 · Filter by author QingCheng24 opened on Oct 23, 2024

• demo can't process in HuggingFace

  SHI-Labs/OneFormer#126 · Filter by author icyriverwolf opened on Oct 23, 2024

• libtorch_cuda_cu.so: cannot open shared object file: No such file or directory

  SHI-Labs/OneFormer#125 · Filter by author wuqun-tju opened on Sep 28, 2024
  1

• Bounding box for each instance with huggingface model

  SHI-Labs/OneFormer#124 · Filter by author prashantmetadome opened on Sep 26, 2024

• Non-deterministic Inference Results After Fine-tuning OneFormer Model

  SHI-Labs/OneFormer#122 · Filter by author PanzaResce opened on Sep 21, 2024

• Swin-L number of epochs trained ?

  SHI-Labs/OneFormer#120 · Filter by author tikr7 opened on Jul 17, 2024

• Is there any way to train only semantic segmentation dataset like ade20k?

  SHI-Labs/OneFormer#119 · Filter by author ET-LONG opened on Jul 15, 2024

• How do I get the best model during training

  SHI-Labs/OneFormer#118 · Filter by author spicycurryz opened on Jun 29, 2024