RuntimeError: stack expect each tensor to be equal size, but got [80000] at entry 0 and [79872] at entry 2

[Zttt0523]

something wrong with the file at ./onsets_and_frames/utils.py and ./onsets_and_frames/train.py

[migperfer]

Thanks for opening this issue. Could you please paste the full stack trace? Otherwise, it is quite hard to debug.

[Zttt0523]

thanks for replying Traceback (most recent calls WITHOUT Sacred internals): File "train.py", line 123, in train for i, batch in zip(loop, cycle(loader)): File "/home/zt/desktop/triad/onsets_and_frames/utils.py", line 11, in cycle for item in iterable: File "/home/zt/anaconda3/envs/triad/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 521, in next data = self._next_data() File "/home/zt/anaconda3/envs/triad/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 561, in _next_data data = self._dataset_fetcher.fetch(index) # may raise StopIteration File "/home/zt/anaconda3/envs/triad/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 52, in fetch return self.collate_fn(data) File "/home/zt/anaconda3/envs/triad/lib/python3.8/site-packages/torch/utils/data/_utils/collate.py", line 74, in default_collate return {key: default_collate([d[key] for d in batch]) for key in elem} File "/home/zt/anaconda3/envs/triad/lib/python3.8/site-packages/torch/utils/data/_utils/collate.py", line 74, in return {key: default_collate([d[key] for d in batch]) for key in elem} File "/home/zt/anaconda3/envs/triad/lib/python3.8/site-packages/torch/utils/data/_utils/collate.py", line 56, in default_collate return torch.stack(batch, 0, out=out) RuntimeError: stack expects each tensor to be equal size, but got [80000] at entry 0 and [79872] at entry 2 i try to do some interpolations,but it seems not work.my MAESTRO dataset is complete,but something wrong with the torch.dataloader and torch.stack due to a mismatched dimensions

[Zttt0523]

thanks again,my torch+cuda is 1.10+cuda111,and run on ubuntu 20.04,and strangely i have entered the training stage,and process stoped at 242/1000000

[Zttt0523]

problem seems solved after i set the sequence_length=79872 instead sample_rate*5 in train.py ,now training process seems starting normally,but i still don't know if this will cause a bad train-loss?

[migperfer]

Having chunks of 5 seconds might not be the best, but the only way of knowing is trying I guess. I'll try to spend some time in this repo and will ping you once I fixed the DataLoader problem.

[migperfer]

Hi! Sorry for the delay, it should work now with commit b432e2e