Mismatch between documentation and code

[V0XNIHILI]

See output shape of audio sample:

And in docs:

However, for all PyTorch audio datasets the shape is (1, timesteps) instead of (timesteps, 1). Should we remove the transpose to the audio here? https://github.com/NeuroBench/neurobench/blob/c01253236d6e84eb5af5cd0c0fdf1bdd84b2bed4/neurobench/datasets/MSWC_dataset.py#L244

[jasonlyik]

Code is correct, documentation is wrong. We want to match the NeuroBench convention of (timesteps, features).

50d855cbd5b361699f48038443a7a5397f5ca855