[BUG] AttributeError: 'ConfigDict' object has no attribute 'test_pipeline'

[pekopoke]

Thanks for your error report and we appreciate it a lot.

Checklist

1. I have searched related issues but cannot get the expected help.
2. I have read the FAQ documentation but cannot get the expected help.
3. The bug has not been fixed in the latest version(dev-1.x).

Describe the bug

----------------------------- Captured stderr call -----------------------------
  Traceback (most recent call last):
    File "/opt/mmflow/mmengine/mmengine/config/config.py", line 51, in __getattr__
      value = super().__getattr__(name)
    File "/opt/conda/lib/python3.7/site-packages/addict/addict.py", line 67, in __getattr__
      return self.__getitem__(item)
    File "/opt/mmflow/mmengine/mmengine/config/config.py", line 47, in __missing__
      raise KeyError(name)
  KeyError: 'test_pipeline'
  
  During handling of the above exception, another exception occurred:
  
  Traceback (most recent call last):
    File "/opt/mmflow/demo/image_demo.py", line 50, in <module>
      main(args)
    File "/opt/mmflow/demo/image_demo.py", line 38, in main
      result = inference_model(model, args.img1, args.img2)
    File "/opt/mmflow/mmflow/apis/inference.py", line 91, in inference_model
      cfg.test_pipeline.remove(t)
    File "/opt/mmflow/mmengine/mmengine/config/config.py", line 53, in __getattr__
      raise AttributeError(f"'{self.__class__.__name__}' object has no "AttributeError: 'ConfigDict' object has no attribute 'test_pipeline'

Reproduction

1. What command or script did you run?

AssertionError: Failed to run image_demo.py

python /opt/mmflow/demo/image_demo.py /opt/mmflow/demo/frame_0001.png /opt/mmflow/demo/frame_0001.png /opt/mmflow/configs/pwcnet/pwcnet_8xb1_slong_flyingchairs-384x448.py pwcnet_8x1_slong_flyingchairs_384x448.pth work_dir

Environment
mmcv 2.0.0rc1
mmflow 1.0.0rc0
mmengine 0.5.0

[LovingThresh]

In inference.py Line 79 - 82，the cfg is changed and the cfg.test_pipeline is replaced by cfg.pipeline.
if isinstance(cfg.test_dataloader, list): cfg = copy.deepcopy(cfg.test_dataloader[0].dataset) else: cfg = copy.deepcopy(cfg.test_dataloader.dataset)

So, the code will run after changing cfg.test_pipeline.remove(t) to cfg.pipeline.remove(t).

[pekopoke]

In inference.py Line 79 - 82，the cfg is changed and the cfg.test_pipeline is replaced by cfg.pipeline. if isinstance(cfg.test_dataloader, list): cfg = copy.deepcopy(cfg.test_dataloader[0].dataset) else: cfg = copy.deepcopy(cfg.test_dataloader.dataset)

So, the code will run after changing cfg.test_pipeline.remove(t) to cfg.pipeline.remove(t).

Thanks a lot， I run after changing cfg.test_pipeline.remove(t) to cfg.pipeline.remove(t) and successfull solved the problem.
I hope the developer could fix code in the dev-1.x.

[manuelmaria101]

Hey.
For some reason my config file was also not working. After finding this trend I traced back the error to the function mmseg/api/utils.py/_prepare_data(). I replaced every instance of test_pipeline by test_dataloader.dataset.pipeline, which should load the test_pipeline that you've setup on the test_dataloader in your config file.

Let me know if this is correct and if it also solved the issue for you.