File size: 5,422 Bytes

704b675

---
tags:
- espnet
- audio
- diarization
language: noinfo
datasets:
- mini_librispeech
license: cc-by-4.0
---

## ESPnet2 DIAR model 

### `jkang/espnet2_mini_librispeech_diar`

This model was trained by jaekookang using mini_librispeech recipe in [espnet](https://github.com/espnet/espnet/).

### Demo: How to use in ESPnet2

```bash
cd espnet
git checkout e08a89e0a43db7fc12bec835c62a000ad10bd417
pip install -e .
cd egs2/mini_librispeech/diar1
./run.sh --skip_data_prep false --skip_train true --download_model jkang/espnet2_mini_librispeech_diar
```

<!-- Generated by scripts/utils/show_diar_result.sh -->
# RESULTS
## Environments
- date: `Tue Feb  8 16:41:16 KST 2022`
- python version: `3.9.7 (default, Sep 16 2021, 13:09:58)  [GCC 7.5.0]`
- espnet version: `espnet 0.10.6a1`
- pytorch version: `pytorch 1.10.1`
- Git hash: `e08a89e0a43db7fc12bec835c62a000ad10bd417`
  - Commit date: `Sun Feb 6 18:54:20 2022 -0500`

## diar_train_diar_raw
### DER
dev_clean_2_ns2_beta2_500
|threshold_median_collar|DER|
|---|---|
|result_th0.3_med11_collar0.0|31.39|
|result_th0.3_med1_collar0.0|31.78|
|result_th0.4_med11_collar0.0|29.99|
|result_th0.4_med1_collar0.0|30.61|
|result_th0.5_med11_collar0.0|29.28|
|result_th0.5_med1_collar0.0|30.19|
|result_th0.6_med11_collar0.0|29.50|
|result_th0.6_med1_collar0.0|30.66|
|result_th0.7_med11_collar0.0|30.90|
|result_th0.7_med1_collar0.0|32.38|

## DIAR config

<details><summary>expand</summary>

```
config: conf/train_diar.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: chunk
output_dir: exp/diar_train_diar_raw
ngpu: 1
seed: 0
num_workers: 1
num_att_plot: 3
dist_backend: nccl
dist_init_method: env://
dist_world_size: null
dist_rank: null
local_rank: 0
dist_master_addr: null
dist_master_port: null
dist_launcher: null
multiprocessing_distributed: false
unused_parameters: false
sharded_ddp: false
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 100
patience: 3
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - valid
    - acc
    - max
keep_nbest_models: 3
nbest_averaging_interval: 0
grad_clip: 5
grad_clip_type: 2.0
grad_noise: false
accum_grad: 2
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: null
use_matplotlib: true
use_tensorboard: true
use_wandb: false
wandb_project: null
wandb_id: null
wandb_entity: null
wandb_name: null
wandb_model_log_interval: -1
detect_anomaly: false
pretrain_path: null
init_param: []
ignore_init_mismatch: false
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/diar_stats_8k/train/speech_shape
- exp/diar_stats_8k/train/spk_labels_shape
valid_shape_file:
- exp/diar_stats_8k/valid/speech_shape
- exp/diar_stats_8k/valid/spk_labels_shape
batch_type: folded
valid_batch_type: null
fold_length:
- 80000
- 800
sort_in_batch: descending
sort_batch: descending
multiple_iterator: false
chunk_length: 200000
chunk_shift_ratio: 0.5
num_cache_chunks: 64
train_data_path_and_name_and_type:
-   - dump/raw/simu/data/train_clean_5_ns2_beta2_500/wav.scp
    - speech
    - sound
-   - dump/raw/simu/data/train_clean_5_ns2_beta2_500/espnet_rttm
    - spk_labels
    - rttm
valid_data_path_and_name_and_type:
-   - dump/raw/simu/data/dev_clean_2_ns2_beta2_500/wav.scp
    - speech
    - sound
-   - dump/raw/simu/data/dev_clean_2_ns2_beta2_500/espnet_rttm
    - spk_labels
    - rttm
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
valid_max_cache_size: null
optim: adam
optim_conf:
    lr: 0.01
scheduler: noamlr
scheduler_conf:
    warmup_steps: 1000
num_spk: 2
init: xavier_uniform
input_size: null
model_conf:
    attractor_weight: 1.0
use_preprocessor: true
frontend: default
frontend_conf:
    fs: 8k
    hop_length: 128
specaug: null
specaug_conf: {}
normalize: global_mvn
normalize_conf:
    stats_file: exp/diar_stats_8k/train/feats_stats.npz
encoder: transformer
encoder_conf:
    input_layer: linear
    num_blocks: 2
    linear_units: 512
    dropout_rate: 0.1
    output_size: 256
    attention_heads: 4
    attention_dropout_rate: 0.0
decoder: linear
decoder_conf: {}
label_aggregator: label_aggregator
label_aggregator_conf: {}
attractor: null
attractor_conf: {}
required:
- output_dir
version: 0.10.6a1
distributed: false
```

</details>



### Citing ESPnet

```BibTex
@inproceedings{watanabe2018espnet,
  author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
  title={{ESPnet}: End-to-End Speech Processing Toolkit},
  year={2018},
  booktitle={Proceedings of Interspeech},
  pages={2207--2211},
  doi={10.21437/Interspeech.2018-1456},
  url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
}




```

or arXiv:

```bibtex
@misc{watanabe2018espnet,
  title={ESPnet: End-to-End Speech Processing Toolkit}, 
  author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
  year={2018},
  eprint={1804.00015},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}
```