File size: 1,046 Bytes
ba35f85
557fb53
 
 
 
51f4763
a8c0792
557fb53
51f4763
 
 
 
557fb53
 
 
 
 
 
 
 
 
 
3a0f0a5
ad4c4e2
 
51f4763
557fb53
 
ba35f85
 
 
 
557fb53
 
 
51f4763
557fb53
ad4c4e2
557fb53
 
 
 
 
 
 
 
 
 
 
3a0f0a5
 
 
 
 
42c4703
3a0f0a5
 
 
557fb53
3a0f0a5
51f4763
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
training_fn: residual.train_residual_dancer
device: mps
seed: 42
dance_ids: &dance_ids
  - BCH
  - BOL
  # - CHA
  - ECS
  - HST
  - LHP
  - NC2
  - JIV
  - QST
  - RMB
  - SFT
  - SLS
  - SMB
  - SWZ
  - TGO
  - VWZ
  - WCS

data_module:
  batch_size: 128
  num_workers: 10
  test_proportion: 0.15

datasets:
  preprocessing.dataset.BestBallroomDataset:
    audio_dir: data/ballroom-songs
    class_list: *dance_ids
    audio_window_jitter: 0.7

  preprocessing.dataset.Music4DanceDataset:
    song_data_path: data/songs_cleaned.csv
    song_audio_path: data/samples
    class_list: *dance_ids
    multi_label: True
    min_votes: 1
    audio_window_jitter: 0.7

model:
  n_channels: 128

feature_extractor:
  mask_count: 0 # Don't mask the data
  snr_mean: 15.0 # Pretty much eliminate the noise
  freq_mask_size: 10
  time_mask_size: 80

trainer:
  log_every_n_steps: 15
  accelerator: gpu
  max_epochs: 50
  min_epochs: 2
  fast_dev_run: False
  # gradient_clip_val: 0.5
  # overfit_batches: 1

training_environment:
  learning_rate: 0.00053