dataloader error

[deepalenYeager]

This is a great piece of work! However, I encountered some issues while trying to train the model using the command “python tools/train.py configs/resnet/rein_resnet50_mask2former_512x512_bs1x4.py” ：

/home/xiaoxu/Rein/rein/models/backbones/dino_layers/swiglu_ffn.py:43: UserWarning: xFormers is available (SwiGLU) warnings.warn("xFormers is available (SwiGLU)") /home/xiaoxu/Rein/rein/models/backbones/dino_layers/attention.py:27: UserWarning: xFormers is available (Attention) warnings.warn("xFormers is available (Attention)") /home/xiaoxu/Rein/rein/models/backbones/dino_layers/block.py:33: UserWarning: xFormers is available (Block) warnings.warn("xFormers is available (Block)") Fail to import ReinsConvNeXt, if you need to use it, please install mmpretrain Traceback (most recent call last): File "/home/xiaoxu/Rein/tools/train.py", line 116, in main() File "/home/xiaoxu/Rein/tools/train.py", line 105, in main runner = Runner.from_cfg(cfg) ^^^^^^^^^^^^^^^^^^^^ File "/home/xiaoxu/.conda/envs/rein/lib/python3.11/site-packages/mmengine/runner/runner.py", line 462, in from_cfg runner = cls( ^^^^ File "/home/xiaoxu/.conda/envs/rein/lib/python3.11/site-packages/mmengine/runner/runner.py", line 342, in init raise ValueError( ValueError: val_dataloader, val_cfg, and val_evaluator should be either all None or not None, but got val_dataloader=None, val_cfg={'type': 'ValLoop'}, val_evaluator=None

I have confirmed that my environment configuration is as follows:

mmcv: 2.1.0 mmdet: 3.3.0 mmengine: 0.10.3 mmsegmentation: 1.2.2 Python: 3.11 PyTorch: 2.0.1 CUDA: 11.7 I am using the Cityscapes dataset.I have tried several methods to resolve this issue but have not found a solution yet. Do you have any suggestions or guidance to help solve this problem?

[w1oves]

It's confused. Can you provide generated entire configs? It should be in the work_dirs.

[deepalenYeager]

Sorry，nothing has been generated under the work_dirs directory, only the directory structure work_dirs/rein_resnet50_mask2former_512x512_bs1x4 has been created。

[w1oves]

Did you modify the base part? The file configs/base/datasets/cityscapes_512x512.py is only used to generate the file configs/base/datasets/dg_citys2acdc_512x512.py. Using the former directly won't make sense.

If you want to train on Cityscapes and test on Cityscapes, I have uploaded the file configs/base/datasets/supervised_cityscapes_512x512.py for you. You only need to replace "../base/datasets/dg_gta_512x512.py" in the base part with this new file.

If you want to train on Cityscapes and test on ACDC, you should use configs/base/datasets/dg_citys2acdc_512x512.py.

[IronmanVsThanos]

当我按照您上述操作后，出现了上述问题。 下面是我的数据集结构

[deepalenYeager]

Thank you so much, I will try it.

[w1oves]

Sorry, i forget to add train_dataloader in configs/_base_/datasets/supervsied_cityscapes_512x512.py. I have updated this.

[IronmanVsThanos]

Thank you so much，but where could i get the dinov2_converted.pth use /tools/converted_models/convet_dinov2?

[IronmanVsThanos]

thank u, it works.