I'm trying to run train.py on exp=base_medium with the ++datamodule.dataset.path= option pointing to my data, but it's failing with the above error. Digging through the code I can only find num_samples in connection with the batch size, which is at the default of 16 (for now). Is there anything I should double-check to get past this error? Stepping through in Debug I can see that the dataset has my paths, so I don't think it's a data problem.
Alternately, is there anything I can just hack to get past this? The only references to num_samples I can find are in the distribution classes in audio-diffusion-pytorch.diffusion.py... maybe just a default there? (Ugly, I know, but possible?)
I'm trying to run
train.pyonexp=base_mediumwith the++datamodule.dataset.path=option pointing to my data, but it's failing with the above error. Digging through the code I can only findnum_samplesin connection with the batch size, which is at the default of 16 (for now). Is there anything I should double-check to get past this error? Stepping through in Debug I can see that the dataset has my paths, so I don't think it's a data problem. Alternately, is there anything I can just hack to get past this? The only references tonum_samplesI can find are in the distribution classes inaudio-diffusion-pytorch.diffusion.py... maybe just a default there? (Ugly, I know, but possible?)Any help greatly appreciated!