ESPnet2 pretrained model, anogkongda/librimix_enh_train_raw_valid.si_snr.ave, fs=8k, lang=en

This model was trained by anogkongda using librimix recipe in espnet.

• Python API

  See https://github.com/espnet/espnet_model_zoo

• Evaluate in the recipe

  git clone https://github.com/espnet/espnet
  cd espnet
  git checkout dcaba2585e28b85c815807165ba9953565ee8694
  pip install -e .
  cd egs2/librimix/enh1
  ./run.sh --skip_data_prep false --skip_train true --download_model anogkongda/librimix_enh_train_raw_valid.si_snr.ave
  

• Results

  # RESULTS
  ## Environments
  - date: `Mon Jan 25 19:16:45 CST 2021`
  - python version: `3.6.3 |Anaconda, Inc.| (default, Nov 20 2017, 20:41:42)  [GCC 7.2.0]`
  - espnet version: `espnet 0.9.7`
  - pytorch version: `pytorch 1.6.0`
  - Git hash: `dcaba2585e28b85c815807165ba9953565ee8694`
    - Commit date: `Thu Jan 21 21:26:59 2021 +0800`
  
  
  ## enh_train_raw
  
  config: ./conf/train.yaml
  sample_rate: 8k
  min_or_max: min
  
  |dataset|STOI|SAR|SDR|SIR|
  |---|---|---|---|---|
  |enhanced_dev|0.845746|11.1029|10.6679|22.6471|
  |enhanced_test|0.846766|10.9166|10.4193|22.0783|

• ENH config

  config: ./conf/train.yaml
  print_config: false
  log_level: INFO
  dry_run: false
  iterator_type: chunk
  output_dir: exp/enh_train_raw
  ngpu: 1
  seed: 0
  num_workers: 4
  num_att_plot: 3
  dist_backend: nccl
  dist_init_method: env://
  dist_world_size: 4
  dist_rank: 0
  local_rank: 0
  dist_master_addr: localhost
  dist_master_port: 48369
  dist_launcher: null
  multiprocessing_distributed: true
  cudnn_enabled: true
  cudnn_benchmark: false
  cudnn_deterministic: true
  collect_stats: false
  write_collected_feats: false
  max_epoch: 200
  patience: 5
  val_scheduler_criterion:
  - valid
  - loss
  early_stopping_criterion:
  - valid
  - loss
  - min
  best_model_criterion:
  -   - valid
      - si_snr
      - max
  -   - valid
      - loss
      - min
  keep_nbest_models: 1
  grad_clip: 5.0
  grad_clip_type: 2.0
  grad_noise: false
  accum_grad: 1
  no_forward_run: false
  resume: true
  train_dtype: float32
  use_amp: false
  log_interval: null
  unused_parameters: false
  use_tensorboard: true
  use_wandb: false
  wandb_project: null
  wandb_id: null
  pretrain_path: null
  init_param: []
  freeze_param: []
  num_iters_per_epoch: null
  batch_size: 16
  valid_batch_size: null
  batch_bins: 1000000
  valid_batch_bins: null
  train_shape_file:
  - exp/enh_stats_8k/train/speech_mix_shape
  - exp/enh_stats_8k/train/speech_ref1_shape
  - exp/enh_stats_8k/train/speech_ref2_shape
  - exp/enh_stats_8k/train/noise_ref1_shape
  valid_shape_file:
  - exp/enh_stats_8k/valid/speech_mix_shape
  - exp/enh_stats_8k/valid/speech_ref1_shape
  - exp/enh_stats_8k/valid/speech_ref2_shape
  - exp/enh_stats_8k/valid/noise_ref1_shape
  batch_type: folded
  valid_batch_type: null
  fold_length:
  - 80000
  - 80000
  - 80000
  - 80000
  sort_in_batch: descending
  sort_batch: descending
  multiple_iterator: false
  chunk_length: 24000
  chunk_shift_ratio: 0.5
  num_cache_chunks: 1024
  train_data_path_and_name_and_type:
  -   - dump/raw/train/wav.scp
      - speech_mix
      - sound
  -   - dump/raw/train/spk1.scp
      - speech_ref1
      - sound
  -   - dump/raw/train/spk2.scp
      - speech_ref2
      - sound
  -   - dump/raw/train/noise1.scp
      - noise_ref1
      - sound
  valid_data_path_and_name_and_type:
  -   - dump/raw/dev/wav.scp
      - speech_mix
      - sound
  -   - dump/raw/dev/spk1.scp
      - speech_ref1
      - sound
  -   - dump/raw/dev/spk2.scp
      - speech_ref2
      - sound
  -   - dump/raw/dev/noise1.scp
      - noise_ref1
      - sound
  allow_variable_data_keys: false
  max_cache_size: 0.0
  max_cache_fd: 32
  valid_max_cache_size: null
  optim: adam
  optim_conf:
      lr: 0.001
      weight_decay: 0
  scheduler: reducelronplateau
  scheduler_conf:
      mode: min
      factor: 0.5
      patience: 1
  init: xavier_uniform
  model_conf:
      loss_type: si_snr
  use_preprocessor: false
  encoder: conv
  encoder_conf:
      channel: 512
      kernel_size: 16
      stride: 8
  separator: tcn
  separator_conf:
      num_spk: 2
      layer: 8
      stack: 3
      bottleneck_dim: 128
      hidden_dim: 512
      kernel: 3
      causal: false
      norm_type: gLN
      nonlinear: relu
  decoder: conv
  decoder_conf:
      channel: 512
      kernel_size: 16
      stride: 8
  required:
  - output_dir
  version: 0.9.7
  distributed: true