| { | |
| "model_type": "sam-audio", | |
| "architectures": [ | |
| "SAMAudioModel" | |
| ], | |
| "num_classes": 8, | |
| "id2label": { | |
| "0": "Amritavarshini", | |
| "1": "Hamsanaadam", | |
| "2": "Kalyani", | |
| "3": "Kharaharapriya", | |
| "4": "Mayamalavagoulai", | |
| "5": "Sindhubhairavi", | |
| "6": "Todi", | |
| "7": "Varali" | |
| }, | |
| "label2id": { | |
| "Amritavarshini": 0, | |
| "Hamsanaadam": 1, | |
| "Kalyani": 2, | |
| "Kharaharapriya": 3, | |
| "Mayamalavagoulai": 4, | |
| "Sindhubhairavi": 5, | |
| "Todi": 6, | |
| "Varali": 7 | |
| }, | |
| "encoder": { | |
| "input_dim": 1, | |
| "hidden_dims": [ | |
| 64, | |
| 128, | |
| 256, | |
| 512 | |
| ], | |
| "kernel_size": 3, | |
| "stride": 2, | |
| "dropout_rate": 0.25, | |
| "use_layer_norm": true, | |
| "use_mel": true, | |
| "n_mels": 80, | |
| "sample_rate": 16000 | |
| }, | |
| "num_segments": 64, | |
| "mask_ratio": 0.0, | |
| "contrastive_temperature": 0.07, | |
| "hidden_size": 512, | |
| "best_val_accuracy": 0.9961977186311787, | |
| "best_epoch": 17 | |
| } |