Upload config.json
Browse files- config.json +100 -0
config.json
ADDED
@@ -0,0 +1,100 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"audio": {
|
3 |
+
"max_len": 1920000,
|
4 |
+
"max_time": 120,
|
5 |
+
"normalize": true,
|
6 |
+
"random_sampling": true,
|
7 |
+
"sample_rate": 16000,
|
8 |
+
"skip_time": false
|
9 |
+
},
|
10 |
+
"augment": {
|
11 |
+
"freq_mask_param": 8,
|
12 |
+
"mixup_alpha": 2.5,
|
13 |
+
"mixup_p": 0.5,
|
14 |
+
"n_freq_masks": 1,
|
15 |
+
"n_time_masks": 2,
|
16 |
+
"time_freq_mask_p": 0.5,
|
17 |
+
"time_mask_param": 8
|
18 |
+
},
|
19 |
+
"dataset": {
|
20 |
+
"test_dataframe": "test.csv",
|
21 |
+
"train_dataframe": "train.csv",
|
22 |
+
"valid_dataframe": "valid.csv"
|
23 |
+
},
|
24 |
+
"environment": {
|
25 |
+
"mixed_precision": true,
|
26 |
+
"num_workers": 2,
|
27 |
+
"seed": 42
|
28 |
+
},
|
29 |
+
"experiment_name": "spectttra_beta-t=120",
|
30 |
+
"logger": {
|
31 |
+
"primary_metric": "f1",
|
32 |
+
"project": "sonics"
|
33 |
+
},
|
34 |
+
"loss": {
|
35 |
+
"label_smoothing": 0.02,
|
36 |
+
"name": "BCEWithLogitsLoss"
|
37 |
+
},
|
38 |
+
"melspec": {
|
39 |
+
"f_max": 8000,
|
40 |
+
"f_min": 20,
|
41 |
+
"hop_length": 512,
|
42 |
+
"n_fft": 2048,
|
43 |
+
"n_mels": 128,
|
44 |
+
"norm": "mean_std",
|
45 |
+
"power": 2,
|
46 |
+
"top_db": 80,
|
47 |
+
"win_length": 2048
|
48 |
+
},
|
49 |
+
"model": {
|
50 |
+
"attn_drop_rate": 0.1,
|
51 |
+
"embed_dim": 384,
|
52 |
+
"f_clip": 3,
|
53 |
+
"input_shape": [
|
54 |
+
128,
|
55 |
+
3744
|
56 |
+
],
|
57 |
+
"mlp_ratio": 2.67,
|
58 |
+
"name": "SpecTTTra",
|
59 |
+
"num_heads": 6,
|
60 |
+
"num_layers": 12,
|
61 |
+
"pe_learnable": true,
|
62 |
+
"pos_drop_rate": 0.1,
|
63 |
+
"pre_norm": true,
|
64 |
+
"proj_drop_rate": 0.0,
|
65 |
+
"resume": null,
|
66 |
+
"t_clip": 5,
|
67 |
+
"use_init_weights": false
|
68 |
+
},
|
69 |
+
"num_classes": 1,
|
70 |
+
"optimizer": {
|
71 |
+
"clip_grad_norm": 5.0,
|
72 |
+
"grad_accum_steps": 1,
|
73 |
+
"momentum": 0.9,
|
74 |
+
"opt": "adamw",
|
75 |
+
"opt_betas": [
|
76 |
+
0.9,
|
77 |
+
0.999
|
78 |
+
],
|
79 |
+
"opt_eps": 1e-08,
|
80 |
+
"weight_decay": 0.05
|
81 |
+
},
|
82 |
+
"scheduler": {
|
83 |
+
"decay_rate": 0.1,
|
84 |
+
"lr": 0.0008,
|
85 |
+
"lr_base": 0.001,
|
86 |
+
"lr_base_scale": "linear",
|
87 |
+
"lr_base_size": 256,
|
88 |
+
"min_lr": 0.0,
|
89 |
+
"sched": "cosine",
|
90 |
+
"warmup_epochs": 5,
|
91 |
+
"warmup_lr": 1e-06
|
92 |
+
},
|
93 |
+
"training": {
|
94 |
+
"batch_size": 128,
|
95 |
+
"epochs": 50
|
96 |
+
},
|
97 |
+
"validation": {
|
98 |
+
"batch_size": 128
|
99 |
+
}
|
100 |
+
}
|