Other Open Access

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

anogkongda

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
Files (14.2 MB)
Name Size
enh_train_raw_valid.si_snr.ave.zip
md5:cffca1922273d84c6e7586c6706b7426
14.2 MB Download
106
151
views
downloads
All versions This version
Views 106105
Downloads 151151
Data volume 2.1 GB2.1 GB
Unique views 102101
Unique downloads 119119

Share

Cite as