| import librosa | |
| ### Audio Hyperparameters ### | |
| sr = 44100 | |
| lf_params = { | |
| 'min_f': librosa.note_to_hz('c0'), | |
| 'max_f': 4100, | |
| 'bins_per_octave': 24, | |
| 'gamma': 20 | |
| } | |
| hf_params = { | |
| 'min_f': 4100, | |
| 'max_f': 16350, | |
| 'bins_per_octave': 96, | |
| 'gamma': 0 | |
| } | |
| ### Network Hyperparameters ### | |
| n_channels = 1 | |
| chunk_size = 512 | |
| frequency_bins = 192 | |
| batch_size = 32 | |
| learning_rate = 0.0001 | |
| epochs = 35 | |
| inference_batch_size = 4 |