| { | |
| "network_config": { | |
| "channels_input": 1, | |
| "channels_output": 1, | |
| "channels_H": 64, | |
| "max_H": 768, | |
| "encoder_n_layers": 8, | |
| "kernel_size": 4, | |
| "stride": 2, | |
| "tsfm_n_layers": 5, | |
| "tsfm_n_head": 8, | |
| "tsfm_d_model": 512, | |
| "tsfm_d_inner": 2048 | |
| }, | |
| "train_config": { | |
| "exp_path": "DNS-large-high", | |
| "log":{ | |
| "directory": "./exp", | |
| "ckpt_iter": "max", | |
| "iters_per_ckpt": 10000, | |
| "iters_per_valid": 500 | |
| }, | |
| "optimization":{ | |
| "n_iters": 250000, | |
| "learning_rate": 2e-4, | |
| "batch_size_per_gpu": 8 | |
| }, | |
| "loss_config":{ | |
| "ell_p": 1, | |
| "ell_p_lambda": 1, | |
| "stft_lambda": 1, | |
| "stft_config":{ | |
| "sc_lambda": 0.5, | |
| "mag_lambda": 0.5, | |
| "band": "high", | |
| "hop_sizes": [50, 120, 240], | |
| "win_lengths": [240, 600, 1200], | |
| "fft_sizes": [512, 1024, 2048] | |
| } | |
| } | |
| }, | |
| "trainset_config": { | |
| "root": "./dns", | |
| "crop_length_sec": 10, | |
| "sample_rate": 16000 | |
| }, | |
| "gen_config":{ | |
| "output_directory": "./exp" | |
| }, | |
| "dist_config": { | |
| "dist_backend": "nccl", | |
| "dist_url": "tcp://localhost:54321" | |
| } | |
| } | |