juice500ml_mls_10h_asr_ssl

Usage

import librosa
from espnet2.bin.asr_inference import Speech2Text

speech2text = Speech2Text.from_pretrained(model_tag="espnet/juice500ml_mls_10h_asr_ssl")
# librosa resamples and mixes to one channel, so any file works; 16000 is
# what nearly every espnet recogniser is trained on - check this model's
# config if its audio is not 16 kHz
speech, rate = librosa.load("audio.wav", sr=16000, mono=True)
text, *_ = speech2text(speech)[0]
print(text)

ESPnet2 ASR model

espnet/juice500ml_mls_10h_asr_ssl

This model was trained by Kwanghee Choi using mls recipe in espnet.

Demo: How to use in ESPnet2

Follow the ESPnet installation instructions if you haven't done that already.

cd espnet
git checkout 29d7cb8453486b9073f729866a8cb3d4a8c203bb
pip install -e .
cd egs2/mls/asr1
./run.sh --skip_data_prep false --skip_train true --download_model espnet/juice500ml_mls_10h_asr_ssl

<!-- Generated by scripts/utils/show_asr_result.sh -->

RESULTS

Environments

• date: Fri Oct 20 23:49:47 EDT 2023
• python version: 3.8.6 (default, Dec 17 2020, 16:57:01) [GCC 10.2.0]
• espnet version: espnet 202308
• pytorch version: pytorch 1.13.1+cu117
• Git hash: 6d5c4220458adc3283838298b549f07dc6aba2ee
  • Commit date: Thu Oct 19 16:01:31 2023 -0400

exp/asr_train_asr_e_branchformer1_wavlm_lr1e-4_raw_bpe150

WER

CER

TER

ASR config

<details><summary>expand</summary>

config: conf/train_asr_e_branchformer1_wavlm_lr1e-4.yaml
print_config: false
log_level: INFO
drop_last_iter: false
dry_run: false
iterator_type: sequence
valid_iterator_type: null
output_dir: exp/asr_train_asr_e_branchformer1_wavlm_lr1e-4_raw_bpe150
ngpu: 1
seed: 2022
num_workers: 4
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: true
sharded_ddp: false
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 18
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - valid
    - acc
    - max
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:
- encoder.encoders
num_iters_per_epoch: null
batch_size: 20
valid_batch_size: null
batch_bins: 1000000
valid_batch_bins: null
train_shape_file:
- exp/asr_stats_raw_bpe150/train/speech_shape
- exp/asr_stats_raw_bpe150/train/text_shape.bpe
valid_shape_file:
- exp/asr_stats_raw_bpe150/valid/speech_shape
- exp/asr_stats_raw_bpe150/valid/text_shape.bpe
batch_type: numel
valid_batch_type: null
fold_length:
- 80000
- 150
sort_in_batch: descending
shuffle_within_batch: false
sort_batch: descending
multiple_iterator: false
chunk_length: 500
chunk_shift_ratio: 0.5
num_cache_chunks: 1024
chunk_excluded_key_prefixes: []
train_data_path_and_name_and_type:
-   - dump/raw/mls_all_train/wav.scp
    - speech
    - sound
-   - dump/raw/mls_all_train/text
    - text
    - text
valid_data_path_and_name_and_type:
-   - dump/raw/mls_all_dev/wav.scp
    - speech
    - sound
-   - dump/raw/mls_all_dev/text
    - text
    - text
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
valid_max_cache_size: null
exclude_weight_decay: false
exclude_weight_decay_conf: {}
optim: adam
optim_conf:
    lr: 0.0001
    weight_decay: 1.0e-06
scheduler: warmuplr
scheduler_conf:
    warmup_steps: 10000
token_list:
- <blank>
- <unk>
- ▁
- s
- a
- e
- o
- i
- t
- u
- n
- l
- r
- m
- d
- g
- en
- y
- f
- ▁a
- p
- ▁p
- er
- z
- ch
- ▁de
- ▁e
- h
- ▁s
- b
- ▁w
- k
- c
- j
- re
- w
- ra
- te
- ▁o
- ar
- ▁t
- an
- ▁z
- ▁i
- ie
- ▁b
- ro
- st
- in
- ł
- or
- v
- ▁g
- 'on'
- é
- ▁di
- li
- ▁d
- ▁la
- de
- ve
- ri
- ▁que
- le
- ▁h
- ta
- ▁ma
- ''''
- ci
- ne
- ▁un
- ▁the
- va
- it
- ▁c
- ▁se
- ▁da
- nd
- ▁no
- la
- do
- ▁m
- ▁k
- ▁po
- ▁in
- ▁le
- ▁he
- ▁si
- to
- ę
- ▁do
- ▁to
- ▁ha
- ce
- ▁en
- is
- ó
- ▁me
- ur
- ▁na
- ▁mi
- ni
- ▁l
- ▁al
- da
- ▁be
- ti
- ▁ca
- me
- ▁vo
- ▁so
- ▁mo
- ą
- ▁ge
- ing
- ▁and
- ż
- q
- ś
- á
- í
- x
- ã
- à
- ü
- ć
- '-'
- ä
- ç
- è
- ß
- ê
- ö
- ñ
- ò
- ú
- ń
- ù
- â
- ô
- ì
- ź
- õ
- î
- û
- ë
- ï
- œ
- æ
- <sos/eos>
init: null
input_size: null
ctc_conf:
    dropout_rate: 0.0
    ctc_type: builtin
    reduce: true
    ignore_nan_grad: null
    zero_infinity: true
joint_net_conf: null
use_preprocessor: true
token_type: bpe
bpemodel: data/token_list/bpe_unigram150/bpe.model
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'
short_noise_thres: 0.5
aux_ctc_tasks: []
frontend: s3prl
frontend_conf:
    frontend_conf:
        upstream: wavlm_large
    download_dir: ./hub
    multilayer_feature: false
    layer: 21
    fs: 16k
specaug: specaug
specaug_conf:
    apply_time_warp: true
    time_warp_window: 5
    time_warp_mode: bicubic
    apply_freq_mask: true
    freq_mask_width_range:
    - 0
    - 27
    num_freq_mask: 2
    apply_time_mask: true
    time_mask_width_ratio_range:
    - 0.0
    - 0.05
    num_time_mask: 5
normalize: utterance_mvn
normalize_conf: {}
model: espnet
model_conf:
    ctc_weight: 0.3
    lsm_weight: 0.1
    length_normalized_loss: false
preencoder: linear
preencoder_conf:
    input_size: 1024
    output_size: 128
encoder: e_branchformer
encoder_conf:
    output_size: 256
    attention_heads: 4
    attention_layer_type: rel_selfattn
    pos_enc_layer_type: rel_pos
    rel_pos_type: latest
    cgmlp_linear_units: 1024
    cgmlp_conv_kernel: 31
    use_linear_after_conv: false
    gate_activation: identity
    num_blocks: 12
    dropout_rate: 0.1
    positional_dropout_rate: 0.1
    attention_dropout_rate: 0.1
    input_layer: conv2d2
    layer_drop_rate: 0.0
    linear_units: 1024
    positionwise_layer_type: linear
    use_ffn: true
    macaron_ffn: true
    merge_conv_kernel: 31
postencoder: null
postencoder_conf: {}
decoder: transformer
decoder_conf:
    attention_heads: 4
    linear_units: 2048
    num_blocks: 6
    dropout_rate: 0.1
    positional_dropout_rate: 0.1
    self_attention_dropout_rate: 0.1
    src_attention_dropout_rate: 0.1
preprocessor: default
preprocessor_conf: {}
required:
- output_dir
- token_list
version: '202308'
distributed: false

</details>

Citing ESPnet

@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:

@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}
}