kevinwang676 commited on
Commit
024b6d3
1 Parent(s): f1a9dc8

Delete config_se.json

Browse files
Files changed (1) hide show
  1. config_se.json +0 -119
config_se.json DELETED
@@ -1,119 +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
- },
52
- "datasets": [
53
- {
54
- "name": "voxceleb2",
55
- "path": "/workspace/scratch/ecasanova/datasets/VoxCeleb/vox2_dev_aac/",
56
- "meta_file_train": null,
57
- "ununsed_speakers": null,
58
- "meta_file_val": null,
59
- "meta_file_attn_mask": "",
60
- "language": "voxceleb"
61
- }
62
- ],
63
- "model_params": {
64
- "model_name": "resnet",
65
- "input_dim": 64,
66
- "use_torch_spec": true,
67
- "log_input": true,
68
- "proj_dim": 512
69
- },
70
- "audio_augmentation": {
71
- "p": 0.5,
72
- "rir": {
73
- "rir_path": "/workspace/store/ecasanova/ComParE/RIRS_NOISES/simulated_rirs/",
74
- "conv_mode": "full"
75
- },
76
- "additive": {
77
- "sounds_path": "/workspace/store/ecasanova/ComParE/musan/",
78
- "speech": {
79
- "min_snr_in_db": 13,
80
- "max_snr_in_db": 20,
81
- "min_num_noises": 1,
82
- "max_num_noises": 1
83
- },
84
- "noise": {
85
- "min_snr_in_db": 0,
86
- "max_snr_in_db": 15,
87
- "min_num_noises": 1,
88
- "max_num_noises": 1
89
- },
90
- "music": {
91
- "min_snr_in_db": 5,
92
- "max_snr_in_db": 15,
93
- "min_num_noises": 1,
94
- "max_num_noises": 1
95
- }
96
- },
97
- "gaussian": {
98
- "p": 0.0,
99
- "min_amplitude": 0.0,
100
- "max_amplitude": 1e-05
101
- }
102
- },
103
- "storage": {
104
- "sample_from_storage_p": 0.5,
105
- "storage_size": 40
106
- },
107
- "max_train_step": 1000000,
108
- "loss": "angleproto",
109
- "grad_clip": 3.0,
110
- "lr": 0.0001,
111
- "lr_decay": false,
112
- "warmup_steps": 4000,
113
- "wd": 1e-06,
114
- "steps_plot_stats": 100,
115
- "num_speakers_in_batch": 100,
116
- "num_utters_per_speaker": 4,
117
- "skip_speakers": true,
118
- "voice_len": 2.0
119
- }