Possible improvement in learning speed by increasing segment_size?

[34j]

Following previous work (Ren et al., 2021; Donahue et al., 2021), we adopt the windowed generator training, a method of generating only a part of raw waveforms to reduce the training time and memory usage during training.

https://ar5iv.labs.arxiv.org/html/2106.06103

This could be different in the case of fine-tuning. Looking for additional validation.

[34j]

Seems not much different