Hugging Face's logo

JYP2024
/
Wedefense_ASV2025_WavLM_Base_Pruning

Audio Classification
English
anti-spoofing
asvspoof5
Model card Files
xet
Community
Wedefense_ASV2025_WavLM_Base_Pruning
File size: 2,155 Bytes 
09c9ee3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
 
data_type: shard
dataloader_args:
  batch_size: 32
  drop_last: true
  num_workers: 6
  pin_memory: false
  prefetch_factor: 8
dataset_args:
  aug_prob: 0.8
  cmvn: true
  cmvn_args:
    norm_mean: true
    norm_var: false
  filter: true
  filter_args:
    max_num_frames: 400
    min_num_frames: 50
  frontend: huggingface
  huggingface_args:
    download_dir: ./s3prl_hub
    frame_length: 20
    frame_shift: 20
    upstream_args:
      name: wavlm_base
      path_or_url: /mnt/matylda6/pengjy/shared_model_weights/lin/wedefense/egs/detection/partialspoof/v15_ssl_mhfa_pruning/convert/wavlm_base_plus_hf/wavlm-base-plus.hf.pth
      pruning_units: conv,head,attlayer,interm,ffnlayer
  num_frms: 150
  resample_rate: 16000
  sample_num_per_epoch: 0
  shuffle: true
  shuffle_args:
    shuffle_size: 2500
  spec_aug: false
  spec_aug_args:
    max_f: 8
    max_t: 10
    num_f_mask: 1
    num_t_mask: 1
    prob: 0.6
  speed_perturb: false
enable_amp: false
exp_dir: exp/v4/MHFA_wavlm_s10
gpus:
- 5
initial_reg_lr: 0.005
log_batch_interval: 100
loss: CrossEntropyLoss
loss_args: {}
margin_scheduler: MarginScheduler
margin_update:
  epoch_iter: 5698
  final_margin: 0.0
  fix_start_epoch: 40
  increase_start_epoch: 20
  increase_type: exp
  initial_margin: 0.0
  update_margin: true
model: SSL_BACKEND_MHFA
model_args:
  compression_dim: 128
  embed_dim: 256
  feat_dim: 768
  feature_grad_mult: 0.1
  head_nb: 32
  nb_layer: 13
model_init: null
noise_data: data/asv5/musan/lmdb
num_avg: 2
num_epochs: 5
optimizer: AdamW
optimizer_args:
  lr: 1.0e-05
  weight_decay: 1.0e-07
original_ssl_num_params: 94422296
projection_args:
  do_lm: false
  easy_margin: false
  embed_dim: 256
  num_class: 2
  project_type: softmax
  scale: 32.0
reg_lr: 0.005
reverb_data: data/asv5/rirs/lmdb/lmdb
save_epoch_interval: 1
scheduler: ExponentialDecrease
scheduler_args:
  epoch_iter: 5698
  final_lr: 1.0e-06
  initial_lr: 1.0e-05
  num_epochs: 5
  scale_ratio: 0.5
  warm_from_zero: true
  warm_up_epoch: 2
seed: 19940716
sparsity_warmup_epochs: 4
target_sparsity: 0.1
train_data: data/asv5/train/shard.list
train_label: data/asv5/train/utt2cls
use_pruning_loss: true