muelea/smplify-xmc

Run demo, but report an error

Closed this issue · 3 comments

code-conquer commented

Thanks for your great works！
When I tried to run your demo, it reported the following error：
RuntimeError: einsum() operands do not broadcast with remapped shapes [original->remapped]: [1, 20]->[1, 1, 1, 20] [10475, 3, 400]->[1, 10475, 3, 400]
Can you help me see what's wrong? Thank you very much!

muelea commented

I will try to reproduce your error, but I need more information for it. Can you please tell me which demo you mean, from where did you download the body model, and from which file this error is raised?

code-conquer commented

I will try to reproduce your error, but I need more information for it. Can you please tell me which demo you mean, from where did you download the body model, and from which file this error is raised?

I download the model from this website:
The specific download is the model under the title of smpl-x model (Download SMPL-X v1.1 (830 MB))
Then I create a new model folder in the root directory of this project and put the downloaded smpl-x model in it. The file structure is as follows:

  • configs
  • data
  • lib
  • model
    • smplx
      • SMPLX_FEMALE.npz
      • SMPLX_MALE.npz
      • SMPLX_NEUTRAL.npz
      • SMPLX_FEMALE.pkl
      • SMPLX_MALE.pkl
      • SMPLX_NEUTRAL.pkl

Finally, I set the path of the model file in the configuration file （configs/fit_smplx_singleview.yaml）
model_folder: model or
mdel_folder: model/smplx/SMPLX_MEUTRAL.npz

python main_singleview.py --config configs/fit_smplx_singleview.yaml \
--dataset mtp_demo \
--input_base_dir data/example_input/singleview/subject1 \
--input_dir_poses data/example_input/presented_poses \
--output_dir data/example_output/singleview/subject1 \
--model_folder model/smplx/SMPLX_NEUTRAL.npz

I set these two paths, but both reported the same error.I need your help. Thank you very much!

woo1 commented

Same error, too.
I updated "model_folder" in "fit_smplx_singleview.yaml" to "models" folder path in smpl-x.

python main_singleview.py --config configs/fit_smplx_singleview.yaml \ --dataset mtp_demo \ --input_base_dir data/example_input/singleview/subject1 \ --input_dir_poses data/example_input/presented_poses \ --output_dir data/example_output/singleview/subject1

Traceback (most recent call last):
File "main_singleview.py", line 243, in
main(**args)
File "main_singleview.py", line 233, in main
**args)
File "/smplify-xmc/lib/core/fit_single_frame.py", line 167, in fit_single_frame
global_orientation_stratey=go_sel_strategy)
File "/smplify-xmc/lib/utils/estimate_global_orient.py", line 74, in best_fit_presented_global_orientation
body = body_model()
File "/smplify-xmc/venv/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/smplify-xmc/venv/lib/python3.7/site-packages/smplx/body_models.py", line 924, in forward
dtype=self.dtype)
File "/smplify-xmc/venv/lib/python3.7/site-packages/smplx/lbs.py", line 179, in lbs
v_shaped = v_template + blend_shapes(betas, shapedirs)
File "/smplify-xmc/venv/lib/python3.7/site-packages/smplx/lbs.py", line 265, in blend_shapes
blend_shape = torch.einsum('bl,mkl->bmk', [betas, shape_disps])
File "/smplify-xmc/venv/lib/python3.7/site-packages/torch/functional.py", line 406, in einsum
return einsum(equation, *_operands)
File "/smplify-xmc/venv/lib/python3.7/site-packages/torch/functional.py", line 408, in einsum
return _VF.einsum(equation, operands) # type: ignore
RuntimeError: einsum() operands do not broadcast with remapped shapes [original->remapped]: [1, 20]->[1, 1, 1, 20] [10475, 3, 400]->[1, 10475, 3, 400]

environment : PyTorch 1.8.1, CUDA 10.2