|
--- |
|
tags: |
|
- espnet |
|
- audio |
|
- self-supervised-learning |
|
language: en |
|
datasets: |
|
- librispeech |
|
license: cc-by-4.0 |
|
--- |
|
|
|
## ESPnet2 SSL model |
|
|
|
### `simpleoier/simpleoier_librispeech_hubert_iter0_train_ssl_torchaudiohubert_base_960h_pretrain_it0_raw` |
|
|
|
This model was trained by simpleoier using librispeech recipe in [espnet](https://github.com/espnet/espnet/). |
|
|
|
### Demo: How to use in ESPnet2 |
|
|
|
Follow the [ESPnet installation instructions](https://espnet.github.io/espnet/installation.html) |
|
if you haven't done that already. |
|
|
|
```bash |
|
cd espnet |
|
git checkout 753f40d61813436d4e76660904d02eaed7a6649e |
|
pip install -e . |
|
cd egs2/librispeech/ssl1 |
|
./run.sh --skip_data_prep false --skip_train true --download_model simpleoier/simpleoier_librispeech_hubert_iter0_train_ssl_torchaudiohubert_base_960h_pretrain_it0_raw |
|
``` |
|
|
|
|
|
|
|
## SSL config |
|
|
|
<details><summary>expand</summary> |
|
|
|
``` |
|
config: conf/tuning/train_ssl_torchaudiohubert_base_960h_pretrain_it0.yaml |
|
print_config: false |
|
log_level: INFO |
|
dry_run: false |
|
iterator_type: sequence |
|
output_dir: exp/hubert_iter0_train_ssl_torchaudiohubert_base_960h_pretrain_it0_raw |
|
ngpu: 1 |
|
seed: 0 |
|
num_workers: 64 |
|
num_att_plot: 3 |
|
dist_backend: nccl |
|
dist_init_method: env:// |
|
dist_world_size: 8 |
|
dist_rank: 0 |
|
local_rank: 0 |
|
dist_master_addr: localhost |
|
dist_master_port: 45091 |
|
dist_launcher: null |
|
multiprocessing_distributed: true |
|
unused_parameters: true |
|
sharded_ddp: false |
|
cudnn_enabled: true |
|
cudnn_benchmark: false |
|
cudnn_deterministic: true |
|
collect_stats: false |
|
write_collected_feats: false |
|
max_epoch: 250 |
|
patience: null |
|
val_scheduler_criterion: |
|
- valid |
|
- loss |
|
early_stopping_criterion: |
|
- valid |
|
- loss |
|
- min |
|
best_model_criterion: |
|
- - valid |
|
- loss |
|
- min |
|
keep_nbest_models: 10 |
|
nbest_averaging_interval: 0 |
|
grad_clip: 5.0 |
|
grad_clip_type: 2.0 |
|
grad_noise: false |
|
accum_grad: 2 |
|
no_forward_run: false |
|
resume: true |
|
train_dtype: float32 |
|
use_amp: true |
|
log_interval: null |
|
use_matplotlib: true |
|
use_tensorboard: true |
|
create_graph_in_tensorboard: false |
|
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: 20 |
|
valid_batch_size: null |
|
batch_bins: 48000000 |
|
valid_batch_bins: null |
|
train_shape_file: |
|
- exp/hubert_iter0_stats_raw/train/speech_shape |
|
- exp/hubert_iter0_stats_raw/train/text_shape.word |
|
valid_shape_file: |
|
- exp/hubert_iter0_stats_raw/valid/speech_shape |
|
- exp/hubert_iter0_stats_raw/valid/text_shape.word |
|
batch_type: numel |
|
valid_batch_type: null |
|
fold_length: |
|
- 80000 |
|
- 400 |
|
sort_in_batch: descending |
|
sort_batch: descending |
|
multiple_iterator: false |
|
chunk_length: 500 |
|
chunk_shift_ratio: 0.5 |
|
num_cache_chunks: 1024 |
|
train_data_path_and_name_and_type: |
|
- - dump/raw/train_960/wav.scp |
|
- speech |
|
- sound |
|
- - dump/raw/train_960/text.km.kmeans_iter0_mfcc_train_960_portion0.1 |
|
- text |
|
- text |
|
valid_data_path_and_name_and_type: |
|
- - dump/raw/dev/wav.scp |
|
- speech |
|
- sound |
|
- - dump/raw/dev/text.km.kmeans_iter0_mfcc_train_960_portion0.1 |
|
- text |
|
- text |
|
allow_variable_data_keys: false |
|
max_cache_size: 0.0 |
|
max_cache_fd: 32 |
|
valid_max_cache_size: null |
|
optim: adam |
|
optim_conf: |
|
lr: 0.0005 |
|
scheduler: warmuplr |
|
scheduler_conf: |
|
warmup_steps: 32000 |
|
token_list: |
|
- '81' |
|
- '5' |
|
- '79' |
|
- '84' |
|
- '27' |
|
- '35' |
|
- '67' |
|
- '56' |
|
- '10' |
|
- '99' |
|
- '24' |
|
- '3' |
|
- '48' |
|
- '8' |
|
- '42' |
|
- '16' |
|
- '32' |
|
- '31' |
|
- '47' |
|
- '43' |
|
- '20' |
|
- '73' |
|
- '49' |
|
- '86' |
|
- '18' |
|
- '64' |
|
- '34' |
|
- '59' |
|
- '95' |
|
- '0' |
|
- '52' |
|
- '44' |
|
- '61' |
|
- '57' |
|
- '30' |
|
- '1' |
|
- '93' |
|
- '6' |
|
- '69' |
|
- '19' |
|
- '7' |
|
- '65' |
|
- '28' |
|
- '89' |
|
- '2' |
|
- '96' |
|
- '91' |
|
- '72' |
|
- '38' |
|
- '78' |
|
- '26' |
|
- '13' |
|
- '39' |
|
- '94' |
|
- '4' |
|
- '88' |
|
- '85' |
|
- '51' |
|
- '82' |
|
- '41' |
|
- '50' |
|
- '21' |
|
- '80' |
|
- '97' |
|
- '87' |
|
- '25' |
|
- '54' |
|
- '12' |
|
- '40' |
|
- '60' |
|
- '29' |
|
- '11' |
|
- '53' |
|
- '71' |
|
- '83' |
|
- '74' |
|
- '68' |
|
- '55' |
|
- '62' |
|
- '76' |
|
- '45' |
|
- '75' |
|
- '92' |
|
- '46' |
|
- '36' |
|
- '66' |
|
- '22' |
|
- '77' |
|
- '23' |
|
- '63' |
|
- '37' |
|
- '58' |
|
- '33' |
|
- '15' |
|
- '17' |
|
- '90' |
|
- '98' |
|
- '14' |
|
- '70' |
|
- '9' |
|
- <unk> |
|
- <sos/eos> |
|
init: null |
|
collate_fn_conf: |
|
label_downsampling: 2 |
|
pad: false |
|
rand_crop: true |
|
input_size: 1 |
|
num_classes: 100 |
|
use_preprocessor: true |
|
token_type: word |
|
bpemodel: null |
|
non_linguistic_symbols: null |
|
cleaner: null |
|
g2p: null |
|
speech_volume_normalize: null |
|
rir_scp: null |
|
rir_apply_prob: 1.0 |
|
noise_scp: null |
|
noise_apply_prob: 1.0 |
|
noise_db_range: '13_15' |
|
pred_masked_weight: 1.0 |
|
pred_nomask_weight: 0.0 |
|
loss_weights: 0.0 |
|
frontend: null |
|
frontend_conf: {} |
|
specaug: null |
|
specaug_conf: {} |
|
normalize: null |
|
normalize_conf: {} |
|
preencoder: null |
|
preencoder_conf: {} |
|
encoder: torchaudio_hubert |
|
encoder_conf: |
|
encoder_projection_dropout: 0.1 |
|
encoder_attention_dropout: 0.1 |
|
encoder_ff_interm_dropout: 0.0 |
|
encoder_dropout: 0.1 |
|
encoder_layer_drop: 0.05 |
|
model: torchaudio |
|
model_conf: {} |
|
required: |
|
- output_dir |
|
- token_list |
|
version: '202209' |
|
distributed: true |
|
``` |
|
|
|
</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} |
|
} |
|
``` |
|
|