ROMP icon indicating copy to clipboard operation
ROMP copied to clipboard

Cannot evaluate on 3DPW

Open kaufManu opened this issue 1 year ago • 1 comments

Hi,

I'm trying to run

python -m romp.test --configs_yml=configs/eval_3dpw_test.yml

but it fails at this line:

INFO:root:3DPW dataset test split total 24423 samples, loading mode vibe, containing 24 video sequence.
Traceback (most recent call last):
  File "miniconda3/envs/ROMP/lib/python3.8/runpy.py", line 194, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "miniconda3/envs/ROMP/lib/python3.8/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "ROMP/romp/test.py", line 204, in <module>
    main()
  File "ROMP/romp/test.py", line 201, in main
    demo.test_eval()
  File "ROMP/romp/test.py", line 27, in test_eval
    MPJPE, PA_MPJPE, eval_results = val_result(self,loader_val=data_loader, evaluation=True)
  File "miniconda3/envs/ROMP/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
    return func(*args, **kwargs)
  File "code/ROMP/romp/eval.py", line 103, in val_result
    outputs = self.network_forward(eval_model, meta_data, self.eval_cfg)
  File "ROMP/romp/base.py", line 121, in network_forward
    outputs = model(meta_data, **self.train_cfg)
AttributeError: 'Demo' object has no attribute 'train_cfg'

This first failed silently as the line is in a try-except block that executes continue if there's an error. I found other issues on Github that report that the evaluation is broken on the newest release, but I didn't unterstand the comments of those threads. Is there a quick way to fix this?

kaufManu avatar May 25 '23 11:05 kaufManu

Hi,

I'm trying to run

python -m romp.test --configs_yml=configs/eval_3dpw_test.yml

but it fails at this line:

INFO:root:3DPW dataset test split total 24423 samples, loading mode vibe, containing 24 video sequence.
Traceback (most recent call last):
  File "miniconda3/envs/ROMP/lib/python3.8/runpy.py", line 194, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "miniconda3/envs/ROMP/lib/python3.8/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "ROMP/romp/test.py", line 204, in <module>
    main()
  File "ROMP/romp/test.py", line 201, in main
    demo.test_eval()
  File "ROMP/romp/test.py", line 27, in test_eval
    MPJPE, PA_MPJPE, eval_results = val_result(self,loader_val=data_loader, evaluation=True)
  File "miniconda3/envs/ROMP/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
    return func(*args, **kwargs)
  File "code/ROMP/romp/eval.py", line 103, in val_result
    outputs = self.network_forward(eval_model, meta_data, self.eval_cfg)
  File "ROMP/romp/base.py", line 121, in network_forward
    outputs = model(meta_data, **self.train_cfg)
AttributeError: 'Demo' object has no attribute 'train_cfg'

This first failed silently as the line is in a try-except block that executes continue if there's an error. I found other issues on Github that report that the evaluation is broken on the newest release, but I didn't unterstand the comments of those threads. Is there a quick way to fix this?

Hi, I also met this problem when evaluation. Do you fix it?

lyk412 avatar Jul 28 '23 06:07 lyk412