When I fuse two perspectives into one video, it works fine during training, but fails during inference.

[wyclearnpy]

My data is like this.

The terminal output is：

Warning: the argument seed shadows a Pipeline constructor argument of the same name.
/opt/dali/dali/operators/reader/nvdecoder/nvdecoder.cc:199: Unable to decode file /home/WYC/datasets/LP_2cam_own/videos/13.mp4
Error executing job with overrides: []
Traceback (most recent call last):
File "/home/WYC/code_repo/lightning-pose/scripts/predict_new_vids.py", line 95, in predict_videos_in_dir
export_predictions_and_labeled_video(
File "/home/WYC/.conda/envs/LP/lib/python3.10/site-packages/typeguard/init.py", line 1033, in wrapper
retval = func(*args, **kwargs)
File "/home/WYC/code_repo/lightning-pose/lightning_pose/utils/scripts.py", line 722, in export_predictions_and_labeled_video
preds_df = predict_single_video(
File "/home/WYC/.conda/envs/LP/lib/python3.10/site-packages/typeguard/init.py", line 1033, in wrapper
retval = func(*args, **kwargs)
File "/home/WYC/code_repo/lightning-pose/lightning_pose/utils/predictions.py", line 425, in predict_single_video
predict_loader = vid_pred_class()
File "/home/WYC/code_repo/lightning-pose/lightning_pose/data/dali.py", line 490, in call
return LitDaliWrapper(pipe, **args[self.train_stage][self.model_type])
File "/home/WYC/code_repo/lightning-pose/lightning_pose/data/dali.py", line 175, in init
super().init(*args, **kwargs)
File "/home/WYC/.conda/envs/LP/lib/python3.10/site-packages/nvidia/dali/plugin/pytorch/init.py", line 224, in init
self._first_batch = DALIGenericIterator.next(self)
File "/home/WYC/.conda/envs/LP/lib/python3.10/site-packages/nvidia/dali/plugin/pytorch/init.py", line 239, in next
outputs = self._get_outputs()
File "/home/WYC/.conda/envs/LP/lib/python3.10/site-packages/nvidia/dali/plugin/base_iterator.py", line 385, in _get_outputs
outputs.append(p.share_outputs())
File "/home/WYC/.conda/envs/LP/lib/python3.10/site-packages/nvidia/dali/pipeline.py", line 1215, in share_outputs
return self._pipe.ShareOutputs()
RuntimeError: Critical error in pipeline:
Error in GPU operator nvidia.dali.fn.readers.video,
which was used in the pipeline definition with the following traceback:

File "/home/WYC/code_repo/lightning-pose/lightning_pose/data/dali.py", line 91, in video_pipe
video = fn.readers.video(

encountered:

Video is too large in at least one dimension.
C++ context: [/opt/dali/dali/operators/reader/nvdecoder/cuvideodecoder.cc:231]
Current pipeline object is no longer valid.

Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.

What is the problem, and how can I solve it？

[themattinthehatt]

@wyclearnpy can you run ffprobe /home/WYC/datasets/LP_2cam_own/videos/13.mp4 from the command line and paste the output here?