# @package _group_

common:
  memory_efficient_fp16: true
  log_format: json
  log_interval: 100
  model_parallel_size: 1

checkpoint:
  no_epoch_checkpoints: true
  best_checkpoint_metric: wer
  save_dir: /esat/spchtemp/scratch/jponcele/selfsupervised_exps/result/finetune_VW_base_all

task:
  _name: audio_pretraining
  data: /users/spraak/jponcele/BenchmarkingSS/data/cgn_phone_10ms_w2v2_all
  normalize: true  #false
  labels: ltr
  segments: true
  max_length: 800000

dataset:
  num_workers: 6
  batch_size: 4
  max_tokens: 32000000
  skip_invalid_size_inputs_valid_test: true
  valid_subset: test
  data_buffer_size: 2

distributed_training:
  ddp_backend: legacy_ddp
  distributed_world_size: 1

criterion:
  _name: ctc
  zero_infinity: true

optimization:
  max_update: 500000
  lr: [0.00003]
  sentence_avg: true
  update_freq: [4]

optimizer:
  _name: adam
  adam_betas: (0.9,0.98)
  adam_eps: 1e-08

lr_scheduler:
  _name: tri_stage
  phase_ratio: [0.1, 0.4, 0.5]
  final_lr_scale: 0.05

model:
  _name: wav2vec_ctc
  w2v_path: /esat/spchtemp/scratch/jponcele/selfsupervised_exps/result/pretrain_w2v2_cgn-unsup-VW_base/checkpoint_74_250000.pt
  apply_mask: true
  mask_prob: 0.65
  mask_channel_prob: 0.5
  mask_channel_length: 64
  layerdrop: 0.1
  activation_dropout: 0.1
  feature_grad_mult: 0.0
  freeze_finetune_updates: 0