Delete config_se.json
Browse files- config_se.json +0 -121
config_se.json
DELETED
|
@@ -1,121 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"model": "speaker_encoder",
|
| 3 |
-
"run_name": "speaker_encoder",
|
| 4 |
-
"run_description": "resnet speaker encoder trained with commonvoice all languages dev and train, Voxceleb 1 dev and Voxceleb 2 dev",
|
| 5 |
-
"epochs": 100000,
|
| 6 |
-
"batch_size": null,
|
| 7 |
-
"eval_batch_size": null,
|
| 8 |
-
"mixed_precision": false,
|
| 9 |
-
"run_eval": true,
|
| 10 |
-
"test_delay_epochs": 0,
|
| 11 |
-
"print_eval": false,
|
| 12 |
-
"print_step": 50,
|
| 13 |
-
"tb_plot_step": 100,
|
| 14 |
-
"tb_model_param_stats": false,
|
| 15 |
-
"save_step": 1000,
|
| 16 |
-
"checkpoint": true,
|
| 17 |
-
"keep_all_best": false,
|
| 18 |
-
"keep_after": 10000,
|
| 19 |
-
"num_loader_workers": 8,
|
| 20 |
-
"num_val_loader_workers": 0,
|
| 21 |
-
"use_noise_augment": false,
|
| 22 |
-
"output_path": "../checkpoints/speaker_encoder/language_balanced/normalized/angleproto-4-samples-by-speakers/",
|
| 23 |
-
"distributed_backend": "nccl",
|
| 24 |
-
"distributed_url": "tcp://localhost:54321",
|
| 25 |
-
"audio": {
|
| 26 |
-
"fft_size": 512,
|
| 27 |
-
"win_length": 400,
|
| 28 |
-
"hop_length": 160,
|
| 29 |
-
"frame_shift_ms": null,
|
| 30 |
-
"frame_length_ms": null,
|
| 31 |
-
"stft_pad_mode": "reflect",
|
| 32 |
-
"sample_rate": 16000,
|
| 33 |
-
"resample": false,
|
| 34 |
-
"preemphasis": 0.97,
|
| 35 |
-
"ref_level_db": 20,
|
| 36 |
-
"do_sound_norm": false,
|
| 37 |
-
"do_trim_silence": false,
|
| 38 |
-
"trim_db": 60,
|
| 39 |
-
"power": 1.5,
|
| 40 |
-
"griffin_lim_iters": 60,
|
| 41 |
-
"num_mels": 64,
|
| 42 |
-
"mel_fmin": 0.0,
|
| 43 |
-
"mel_fmax": 8000.0,
|
| 44 |
-
"spec_gain": 20,
|
| 45 |
-
"signal_norm": false,
|
| 46 |
-
"min_level_db": -100,
|
| 47 |
-
"symmetric_norm": false,
|
| 48 |
-
"max_norm": 4.0,
|
| 49 |
-
"clip_norm": false,
|
| 50 |
-
"stats_path": null,
|
| 51 |
-
"do_rms_norm": true,
|
| 52 |
-
"db_level": -27.0
|
| 53 |
-
},
|
| 54 |
-
"datasets": [
|
| 55 |
-
{
|
| 56 |
-
"name": "voxceleb2",
|
| 57 |
-
"path": "/workspace/scratch/ecasanova/datasets/VoxCeleb/vox2_dev_aac/",
|
| 58 |
-
"meta_file_train": null,
|
| 59 |
-
"ununsed_speakers": null,
|
| 60 |
-
"meta_file_val": null,
|
| 61 |
-
"meta_file_attn_mask": "",
|
| 62 |
-
"language": "voxceleb"
|
| 63 |
-
}
|
| 64 |
-
],
|
| 65 |
-
"model_params": {
|
| 66 |
-
"model_name": "resnet",
|
| 67 |
-
"input_dim": 64,
|
| 68 |
-
"use_torch_spec": true,
|
| 69 |
-
"log_input": true,
|
| 70 |
-
"proj_dim": 512
|
| 71 |
-
},
|
| 72 |
-
"audio_augmentation": {
|
| 73 |
-
"p": 0.5,
|
| 74 |
-
"rir": {
|
| 75 |
-
"rir_path": "/workspace/store/ecasanova/ComParE/RIRS_NOISES/simulated_rirs/",
|
| 76 |
-
"conv_mode": "full"
|
| 77 |
-
},
|
| 78 |
-
"additive": {
|
| 79 |
-
"sounds_path": "/workspace/store/ecasanova/ComParE/musan/",
|
| 80 |
-
"speech": {
|
| 81 |
-
"min_snr_in_db": 13,
|
| 82 |
-
"max_snr_in_db": 20,
|
| 83 |
-
"min_num_noises": 1,
|
| 84 |
-
"max_num_noises": 1
|
| 85 |
-
},
|
| 86 |
-
"noise": {
|
| 87 |
-
"min_snr_in_db": 0,
|
| 88 |
-
"max_snr_in_db": 15,
|
| 89 |
-
"min_num_noises": 1,
|
| 90 |
-
"max_num_noises": 1
|
| 91 |
-
},
|
| 92 |
-
"music": {
|
| 93 |
-
"min_snr_in_db": 5,
|
| 94 |
-
"max_snr_in_db": 15,
|
| 95 |
-
"min_num_noises": 1,
|
| 96 |
-
"max_num_noises": 1
|
| 97 |
-
}
|
| 98 |
-
},
|
| 99 |
-
"gaussian": {
|
| 100 |
-
"p": 0.0,
|
| 101 |
-
"min_amplitude": 0.0,
|
| 102 |
-
"max_amplitude": 1e-05
|
| 103 |
-
}
|
| 104 |
-
},
|
| 105 |
-
"storage": {
|
| 106 |
-
"sample_from_storage_p": 0.5,
|
| 107 |
-
"storage_size": 40
|
| 108 |
-
},
|
| 109 |
-
"max_train_step": 1000000,
|
| 110 |
-
"loss": "angleproto",
|
| 111 |
-
"grad_clip": 3.0,
|
| 112 |
-
"lr": 0.0001,
|
| 113 |
-
"lr_decay": false,
|
| 114 |
-
"warmup_steps": 4000,
|
| 115 |
-
"wd": 1e-06,
|
| 116 |
-
"steps_plot_stats": 100,
|
| 117 |
-
"num_speakers_in_batch": 100,
|
| 118 |
-
"num_utters_per_speaker": 4,
|
| 119 |
-
"skip_speakers": true,
|
| 120 |
-
"voice_len": 2.0
|
| 121 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|