Repository Issue Activity (beta)

#361 With `--use-amp`, the FGL/DDF losses run in fp16 (criterion under `autocast(enabled=False)` on fp16 outputs) — DDF reads ≈ 5× its fp32 value

#358 Inconsistent preprocessing between PyTorch (Stretch) and ONNX (Letterbox) scripts causes accuracy drop and class flickering

#357 License status of the published D-FINE pretrained weights (follow-up to #162)

#173 Instance Segmentation

#202 我在通过debug的方式看代码的时候，为什么pred_corners的值每一层都是0？

#37 关于VisDrone2019数据集训练的精度问题

#348 背景类图片

#94 Extending model to hyperspectral images?

#57 Author, could you please explain why the model is not training?

#42 能帮忙看一下为什么报错吗

#226 Why does FGL loss fluctuate around 1

#342 Discrepancy in Parameter Count Between get_info.py and train.py --test-only for D-FINE-S Model

#347 Cannot stop training

#352 每次结果偏差很大

#354 Security: unsafe YAML config loading enables arbitrary code execution (CWE-502)

#333 d-fine模型这个标题怎么理解？为什么会取这个名字

#316 训练到第26个epoch卡死

#350 Wrapper easy to use

#18 TensorRT Dynamic shape

#335 Convert D-Fine to CoreML

#179 Has anyone been converted the model to INT8 successfully?

#145 Errors caused by replacing the decoder part：module_kwargs[k] = create(_cfg['_name'], global_cfg) KeyError: '_name'

#197 多卡训练中断（Multi-GPU training is interrupted, consistently failing around epoch 190.）

#349 TensorRT benchmarking parameters on T4 GPU

#170 NotImplementedError