Skip to content

GitLab

Switch branch/tag
Find file Normal viewHistoryPermalink
w2vu2.yaml 3.1 KB
Newer Older
huaerkl's avatar
v1.0  
huaerkl committed
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
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154 
# @package _group_

common:
  fp16: false
  fp16_no_flatten_grads: true
  log_format: json
  log_interval: 100
  tensorboard_logdir: tb
  reset_logging: false
  suppress_crashes: false

checkpoint:
  save_interval: 1000
  save_interval_updates: 1000
  no_epoch_checkpoints: true
  best_checkpoint_metric: weighted_lm_ppl
  save_dir: .

distributed_training:
  distributed_world_size: 1

task:
  _name: unpaired_audio_text
  data: ???
  text_data: ???
  labels: phn
  sort_by_length: false
  unfiltered: false
  max_length: null
  append_eos: false
  kenlm_path: ???
  aux_target_postfix: km

dataset:
  num_workers: 6
  batch_size: 160
  skip_invalid_size_inputs_valid_test: true
  valid_subset: valid
  validate_interval: 1000
  validate_interval_updates: 1000

criterion:
  _name: model
  log_keys:
    - accuracy_dense
    - accuracy_token
    - temp
    - code_ppl

optimization:
  max_update: 150000
  clip_norm: 5.0
  lr: [0]

optimizer:
  _name: composite
  groups:
    generator:
      lr: [0.00005]
      lr_float: null
      optimizer:
        _name: adam
        adam_betas: [0.5,0.98]
        adam_eps: 1e-06
        weight_decay: 0
        amsgrad: false
      lr_scheduler:
        _name: fixed
        warmup_updates: 0
    discriminator:
      lr: [ 0.0003 ]
      lr_float: null
      optimizer:
        _name: adam
        adam_betas: [0.5,0.98]
        adam_eps: 1e-06
        weight_decay: 0.0001
        amsgrad: false
      lr_scheduler:
        _name: fixed
        warmup_updates: 0

lr_scheduler: pass_through

model:
  _name: wav2vec_u

  discriminator_dim: 384
  discriminator_depth: 2
  discriminator_kernel: 8
  discriminator_linear_emb: false
  discriminator_causal: true
  discriminator_max_pool: false
  discriminator_act_after_linear: false
  discriminator_dropout: 0.0
  discriminator_weight_norm: false

  generator_stride: 3
  generator_kernel: 9
  generator_bias: false
  generator_dropout: 0.1
  generator_batch_norm: 30
  generator_residual: true

  smoothness_weight: 1.5
  smoothing: 0
  smoothing_one_sided: false
  gumbel: false
  hard_gumbel: false
  gradient_penalty: 1.0
  code_penalty: 3.0
  temp: [ 2,0.1,0.99995 ]
  input_dim: 1024
  mmi_weight: 0.5
  target_dim: 64

  segmentation:
    type: JOIN
    mean_pool_join: false
    remove_zeros: false


hydra:
  job:
    config:
      override_dirname:
        kv_sep: ':'
        item_sep: '__'
        exclude_keys:
          - run_config
          - distributed_training.distributed_port
          - common.user_dir
          - task.data
          - task.kenlm_path
          - task.text_data
          - model.generator_layers
          - task.labels
          - task.force_model_seed
  sweep:
    dir: /checkpoint/${env:USER}/${env:PREFIX}/${hydra.job.config_name}/${hydra.job.override_dirname}
    subdir: ${hydra.job.num}
  launcher:
    submitit_folder: ${hydra.sweep.dir}
    timeout_min: 3000
    cpus_per_task: 10
    gpus_per_node: 1
    tasks_per_node: 1
    mem_gb: 120
    nodes: 1
    name: ${env:PREFIX}_${hydra.job.config_name}
    partition: devlab,learnlab,learnfair,scavenge
    comment: intern_endding_soon
    constraint: volta32gb
    max_num_timeout: 30