AttributeError: can't set attribute - running sample.conditional_synthesis

[icedwater]

I can't seem to get the sample command to work on my end even after preparing the HumanML3D dataset. Am I missing anything obvious? The console output is below.

$ python -m sample.conditional_synthesis --model_path ./save/randomframes/model000750000.pt --edit_mode benchmark_sparse --transition_length 5 --num_samples 10 --num_repetitions 3 --text_prompt "a person throws a ball"
Loading dataset...
====DatasetConfig(name='humanml', batch_size=1, num_frames=196, split='test', hml_mode='train', use_abs3d=True, traject_only=False, use_random_projection=False, random_projection_scale=10.0, augment_type='none', std_scale_shift=[1.0, 0.0], drop_redundant=False)====
====humanml====
Reading ././dataset/humanml_opt.txt
WARNING: max_motion_length is set to 196
Loading dataset t2m ...
mode = train
t2m dataset aug: none std_scale_shift: [1.0, 0.0]
t2m dataset drop redundant information: False
--- Using small data for debugging ---
100%|███████████| 30/30 [00:00<00:00, 5428.82it/s]
Pointer Pointing at 0
Creating model and diffusion...
Using UNET with lantent dim:  512  and mults:  [2, 2, 2, 2]
dims:  [263, 1024, 1024, 1024, 1024] mults:  [2, 2, 2, 2]
[ models/temporal ] Channel dimensions: [(263, 1024), (1024, 1024), (1024, 1024), (1024, 1024)]
EMBED TEXT
Loading CLIP...
Loading checkpoints from [./save/randomframes/model000750000.pt]...
loading avg model
### Start sampling [repetitions #0]
100%|█████████████████████| 1000/1000 [00:35<00:00, 28.36it/s]
created 1 samples
### Start sampling [repetitions #1]
100%|█████████████████████| 1000/1000 [00:34<00:00, 28.69it/s]
created 2 samples
### Start sampling [repetitions #2]
100%|█████████████████████| 1000/1000 [00:33<00:00, 29.68it/s]
created 3 samples
saving results file to [save/results/randomframes/condsamples000750000__benchmark_sparse_T=5_CI=0_CRG=0_KGP=1.0_seed10_a_person_throws_a_ball/results.npy]
[(0) "Input Motion" | -> input_motion00.mp4]
Traceback (most recent call last):
  File "/home/iced/anaconda3/lib/python3.9/runpy.py", line 197, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/home/iced/anaconda3/lib/python3.9/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/home/iced/projects/CondMDI/sample/conditional_synthesis.py", line 314, in <module>
    main()
  File "/home/iced/projects/CondMDI/sample/conditional_synthesis.py", line 282, in main
    plot_conditional_samples(motion=all_motions,
  File "/home/iced/projects/CondMDI/data_loaders/humanml/utils/plotting.py", line 83, in plot_conditional_samples
    plot_3d_motion(animation_save_path, skeleton, gt_motion, title=caption,
  File "/home/iced/projects/CondMDI/data_loaders/humanml/utils/plot_script.py", line 242, in plot_3d_motion
    ani.save(save_path, fps=fps)
  File "/home/iced/.venv/condmdi/lib/python3.9/site-packages/matplotlib/animation.py", line 1072, in save
    anim._init_draw()  # Clear the initial frame
  File "/home/iced/.venv/condmdi/lib/python3.9/site-packages/matplotlib/animation.py", line 1727, in _init_draw
    self._draw_frame(frame_data)
  File "/home/iced/.venv/condmdi/lib/python3.9/site-packages/matplotlib/animation.py", line 1746, in _draw_frame
    self._drawn_artists = self._func(framedata, *self._args)
  File "/home/iced/projects/CondMDI/data_loaders/humanml/utils/plot_script.py", line 184, in update
    ax.lines = []
AttributeError: can't set attribute

[icedwater]

Also, I'm not sure why the first part tries to run using a conda environment but the error comes from my venv - this appears even after I've repeatedly done conda deactivate.

[icedwater]

This workaround helped me to get the video generated, though now the frames aren't cleared and the generated motion is messy. (Edit: it was because I forgot to add the ax.clear() line. Now the frames are cleared.)

[setarehc]

Yes, this is related to the matplotlib version you're using. I use matplotlib 3.1.3 and the code works fine.
Glad to hear the workaround suggested for the issue you posted, resolved the problem.