zl389 commited on
Commit
bd36143
·
verified ·
1 Parent(s): fcf54e3

Upload train.yaml

Browse files
Files changed (1) hide show
  1. train.yaml +81 -0
train.yaml ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ seed: 24
2
+ __set_seed: !apply:deeplab.utils.misc.set_random_seed [!ref <seed>]
3
+
4
+ use_amp: true
5
+ use_gradient_clipping: true
6
+ gradient_accumulation: 1
7
+ cudnn_benchmark: false
8
+
9
+ optimizer: !name:torch.optim.AdamW
10
+ lr: 0.00001
11
+ weight_decay: 0.0001
12
+
13
+ scheduler_lmft: !name:deeplab.core.scheduler.WarmupCosineScheduler
14
+ min_lr: 0.000005
15
+ max_lr: 0.00001
16
+ warmup_epoch: 0
17
+ fix_epoch: 1
18
+
19
+ num_epochs: 2
20
+ max_iters_per_epoch: null
21
+ batch_size: 32
22
+ valid_batch_size: 1
23
+ training_loop: 1
24
+ num_workers: 16
25
+ output_dir: results/audio2vec_new
26
+
27
+ items_save: True
28
+ item_save_steps: 500
29
+
30
+ dur_range: [5, 6]
31
+ max_valid_dur: 60
32
+ speed_perturbation: []
33
+ data_aug: false
34
+ embd_dim: 256
35
+
36
+ sample_rate: 16000
37
+
38
+ corpus_dir: /work/zl389/AudioData
39
+
40
+ musan_path: !apply:os.path.join [!ref <corpus_dir>, 'musan']
41
+ rirs_path: !apply:os.path.join [!ref <corpus_dir>, 'rirs_noise']
42
+
43
+ train_data:
44
+ - !name:deeplab.utils.corpus.load_audio_corpus
45
+ ['/work/zl389/workspace/LLM_ASV/data/vox2dev', ['dev']] # 5994
46
+
47
+ valid_data:
48
+ - scp_path: '/work/zl389/workspace/LLM_ASV/data/test_vox/vox1-o/wav_copy.scp'
49
+ trial_path: '/work/zl389/workspace/LLM_ASV/data/test_vox/vox1-o/trials'
50
+
51
+ # peft_config: !apply:deeplab.pretrained.audio2vec.api.create_lora_config
52
+ # model_type: 'w2v-bert'
53
+ # r: 64
54
+ # lora_alpha: 128
55
+ # target_modules: ["linear_q", "linear_v"]
56
+ # lora_dropout: 0.0
57
+ # bias: 'none'
58
+ peft_config: null
59
+
60
+ spk_model: !new:local.spk_model.Audio2Vec_based_Adapter
61
+ model_name: 'facebook/w2v-bert-2.0'
62
+ frozen_encoder: false
63
+ bnb_config: null
64
+ peft_config: !ref <peft_config>
65
+ encoder_config: 'config_prune_tea.json'
66
+ n_mfa_layers: -1
67
+ pooling_layer: 'ASP'
68
+ embd_dim: !ref <embd_dim>
69
+ adapter_dim: 128
70
+ dropout: 0.0
71
+
72
+
73
+ classifier: !new:local.spk_classifier.ArcFace
74
+ in_features: !ref <embd_dim>
75
+ out_features: 5994
76
+ s: 32
77
+ m: 0.5
78
+
79
+ modules:
80
+ spk_model: !ref <spk_model>
81
+ classifier: !ref <classifier>