| { | |
| "model_type": "musicgen", | |
| "text_encoder": { | |
| "model_type": "t5", | |
| "name_or_path": "t5-base", | |
| "type": "T5EncoderModel", | |
| "config": { | |
| "vocab_size": 32128, | |
| "d_model": 1024, | |
| "num_layers": 12, | |
| "num_heads": 16, | |
| "dropout_rate": 0.1 | |
| } | |
| }, | |
| "audio_encoder": { | |
| "model_type": "wav2vec2", | |
| "name_or_path": "facebook/wav2vec2-large", | |
| "type": "AudioEncoder", | |
| "config": { | |
| "sample_rate": 32000, | |
| "num_channels": 1, | |
| "embedding_size": 512 | |
| } | |
| }, | |
| "decoder": { | |
| "model_type": "transformer", | |
| "name_or_path": "facebook/musicgen-large", | |
| "type": "TransformerDecoder", | |
| "config": { | |
| "d_model": 1024, | |
| "num_heads": 16, | |
| "num_layers": 24, | |
| "dropout_rate": 0.1 | |
| } | |
| }, | |
| "training": { | |
| "batch_size": 16, | |
| "num_epochs": 100, | |
| "learning_rate": 0.0001, | |
| "weight_decay": 0.01, | |
| "gradient_clipping": 1.0 | |
| }, | |
| "generation": { | |
| "sample_rate": 32000, | |
| "audio_format": "wav", | |
| "num_samples": 5, | |
| "max_duration": 30.0, | |
| "temperature": 1.0, | |
| "top_k": 250, | |
| "top_p": 0.9 | |
| }, | |
| "logging": { | |
| "log_tensorboard": true, | |
| "log_wandb": true, | |
| "wandb_project": "music_generation", | |
| "log_updates": 10 | |
| }, | |
| "hardware": { | |
| "device": "cuda", | |
| "num_gpus": 4 | |
| } | |
| } |