Closed dylanrandle closed 4 years ago
It only has the original paper baseline, so it may not be worth using DARTS on it. Getting better results on a dataset that has been manually explored by many people would probably be more meaningful. Let's keep this one in mind if it can be useful to explore architecture transferring / sparsity comparison etc.
Train DARTS with Music Dataset, try to get state-of-art