slueted_owsm_summ

ESPnet2 ASR model

espnet/slueted_owsm_summ

This model was trained by “siddhu001” using slue-ted 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 e23ef85f0b3116ad5c60d0833f186da0deec0734
pip install -e .
cd egs2/slue-ted/slu1
./run.sh --skip_data_prep false --skip_train true --download_model espnet/slueted_owsm_summ

RESULT 23.72001537034864 5.0727273644425015 16.49858244502415 22.432552212134674 83.55819195438502

ASR config

<details><summary>expand</summary>

config: conf//train_asr_owsm_weighted_dropout.yaml
print_config: false
log_level: INFO
drop_last_iter: false
dry_run: false
iterator_type: sequence
valid_iterator_type: null
output_dir: exp/slu_train_asr_owsm_weighted_dropout_raw_en_bpe500_sp
ngpu: 1
seed: 2022
num_workers: 2
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: 55179
dist_launcher: null
multiprocessing_distributed: true
unused_parameters: false
sharded_ddp: false
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 25
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: 1
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: 100
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
use_lora: false
save_lora_only: true
lora_conf: {}
pretrain_path: null
init_param:
- /scratch/bbjs/arora1/espnet_slue_PR/espnet/egs2/tedlium3/asr1/exp/asr_train_asr_owsm_weighted_dropout_raw_en_bpe500/valid.acc.ave_10best.pth:::ctc
ignore_init_mismatch: false
freeze_param:
- encoder
num_iters_per_epoch: null
batch_size: 20
valid_batch_size: null
batch_bins: 12000000
valid_batch_bins: null
train_shape_file:
- exp/slu_stats_raw_en_bpe500_sp/train/speech_shape
- exp/slu_stats_raw_en_bpe500_sp/train/text_shape.bpe
valid_shape_file:
- exp/slu_stats_raw_en_bpe500_sp/valid/speech_shape
- exp/slu_stats_raw_en_bpe500_sp/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: []
chunk_default_fs: null
train_data_path_and_name_and_type:
-   - dump/raw/train_sp/wav.scp
    - speech
    - kaldi_ark
-   - dump/raw/train_sp/text
    - text
    - text
valid_data_path_and_name_and_type:
-   - dump/raw/devel/wav.scp
    - speech
    - kaldi_ark
-   - dump/raw/devel/text
    - text
    - text
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
allow_multi_rates: false
valid_max_cache_size: null
exclude_weight_decay: false
exclude_weight_decay_conf: {}
optim: adam
optim_conf:
    lr: 0.002
    weight_decay: 1.0e-06
scheduler: warmuplr
scheduler_conf:
    warmup_steps: 5000
token_list:
- <blank>
- <unk>
- '[sep]'
- '&quot;'
- s
- ▁
- ▁the
- ','
- t
- d
- ▁a
- .
- ing
- o
- e
- ▁to
- a
- ▁and
- y
- n
- ▁of
- r
- ▁in
- u
- i
- m
- p
- c
- er
- g
- l
- al
- re
- ed
- b
- ''''
- ar
- k
- in
- f
- ▁"
- le
- 'on'
- v
- or
- th
- '-'
- ▁c
- en
- ▁f
- ▁--
- ▁we
- ▁for
- ▁how
- ly
- ▁re
- se
- ▁that
- es
- w
- ic
- st
- ▁w
- ▁be
- ri
- an
- ra
- ve
- ce
- ur
- ▁by
- ▁it
- li
- ▁de
- '?'
- it
- ch
- ent
- ▁is
- ter
- el
- ▁on
- ▁e
- ▁he
- ▁co
- ▁an
- ▁ma
- ▁st
- ll
- ▁with
- ▁can
- il
- ▁you
- ▁us
- ation
- te
- ▁this
- ▁b
- ▁do
- ▁g
- me
- ▁what
- ck
- ▁from
- ate
- ▁p
- z
- la
- ▁mo
- ▁di
- ive
- mp
- ▁talk
- ity
- vi
- ta
- at
- ge
- ▁tr
- ▁she
- ▁our
- ▁pa
- ci
- et
- h
- ▁su
- ver
- ▁world
- pe
- ▁about
- ▁me
- ▁so
- and
- ▁con
- tion
- de
- ir
- ▁her
- im
- ':'
- ▁his
- ies
- ▁po
- ▁are
- ect
- lo
- ▁your
- un
- ist
- hi
- ▁mi
- x
- id
- ment
- ol
- ul
- ti
- ne
- qu
- ▁but
- ▁ca
- ▁fa
- ▁as
- ▁un
- ers
- ight
- ▁says
- '0'
- ng
- op
- '1'
- ▁k
- ad
- j
- ma
- ▁pro
- ▁work
- ▁ba
- ▁share
- ▁new
- ▁more
- ▁vi
- ▁sa
- ▁at
- ▁la
- ut
- bi
- sion
- ▁ho
- na
- act
- age
- ke
- if
- ▁bo
- ▁br
- ▁ha
- ▁no
- co
- ▁lo
- mi
- ▁make
- ▁people
- ▁why
- ant
- ▁their
- ▁i
- ▁life
- ▁all
- ting
- ▁human
- ▁have
- om
- )
- ▁(
- ▁help
- ▁ted
- wa
- sh
- ▁da
- ▁le
- ▁out
- ph
- ical
- ▁way
- ff
- ▁ro
- able
- ▁some
- est
- ure
- em
- ho
- ▁ex
- gen
- ha
- ia
- ine
- ▁into
- ca
- ▁was
- ▁who
- ther
- ▁they
- ow
- he
- ▁one
- ▁when
- form
- ▁pre
- ni
- ▁could
- ▁like
- ▁per
- ▁up
- ance
- com
- ▁go
- ion
- tor
- ▁fe
- ▁ra
- ▁or
- ▁en
- ▁change
- tic
- ▁every
- ▁jo
- ence
- ▁not
- ▁art
- one
- use
- ous
- ▁plan
- ▁music
- ▁exp
- und
- ▁ne
- um
- ative
- pp
- ▁need
- tro
- directed
- ▁learn
- ▁narrate
- ▁has
- lar
- '].'
- man
- ▁car
- ▁future
- ▁real
- ▁time
- ize
- ▁live
- ber
- ▁mar
- ▁ga
- ▁take
- ▁dr
- ful
- ▁get
- ▁shows
- day
- ▁cha
- ▁than
- ▁know
- ian
- ▁see
- ▁just
- '2'
- ▁other
- old
- ▁design
- ▁chi
- ▁build
- ious
- ▁most
- ▁si
- ▁will
- ▁power
- ▁think
- port
- ▁over
- ▁ja
- ish
- ▁climate
- ▁sha
- ▁through
- less
- '3'
- ▁my
- ▁where
- ▁global
- ▁health
- ▁pri
- ▁20
- ▁story
- gu
- ugh
- ▁create
- ▁look
- ▁trans
- ▁har
- ▁even
- ▁part
- ▁years
- ▁lead
- side
- low
- long
- ▁technolog
- ness
- '5'
- ▁call
- ▁sc
- ▁system
- '9'
- line
- ▁brain
- ▁data
- ▁own
- ition
- ▁explains
- ▁tell
- ▁explore
- ▁start
- ▁ru
- ▁which
- ▁anderson
- ▁find
- ▁hu
- ▁women
- ▁better
- ▁idea
- ▁history
- ▁research
- ▁science
- ism
- ▁first
- ▁grow
- ▁right
- clu
- ▁space
- ▁develop
- ▁problem
- ▁two
- ▁earth
- ologist
- ▁many
- ▁should
- ▁three
- ▁fellow
- ▁social
- ▁africa
- ▁...
- '4'
- ▁addis
- ▁powerful
- ▁found
- ▁under
- ▁understand
- ▁after
- ▁stories
- ▁around
- ▁personal
- ▁project
- ▁between
- ▁question
- ▁play
- ▁scientist
- ▁happen
- ▁good
- ▁produc
- ▁experience
- ▁step
- ▁america
- '8'
- ▁great
- ▁down
- ▁high
- ▁would
- ▁turn
- ▁surprising
- ▁imagin
- ▁teach
- cross
- ▁place
- ▁medic
- ▁million
- ▁things
- '7'
- ▁reveal
- ▁without
- ▁challenge
- ▁next
- ▁each
- ▁studio
- organ
- '6'
- ▁business
- ▁much
- ▁show
- ▁conversation
- ▁energy
- ▁school
- ▁ocean
- ▁while
- source
- ization
- ▁break
- ▁robot
- ▁disease
- ▁behind
- ability
- ▁team
- ▁chris
- ▁become
- ▁solution
- ▁protect
- ▁collect
- ▁different
- ▁those
- ▁connect
- ▁architect
- ▁language
- ▁simple
- ▁solve
- ▁before
- ▁community
- ▁country
- ▁secret
- ▁keep
- ▁food
- ▁thought
- ▁discover
- ▁environment
- ▁government
- ▁public
- ;
- '!'
- /
- q
- '%'
- '@'
- ']'
- +
- '&'
- '|'
- _
- (
- '"'
- $
- '*'
- '='
- '['
- '`'
- <sos/eos>
transcript_token_list: null
two_pass: false
pre_postencoder_norm: false
init: null
input_size: null
ctc_conf:
    dropout_rate: 0.0
    ctc_type: builtin
    reduce: true
    ignore_nan_grad: null
    zero_infinity: true
    brctc_risk_strategy: exp
    brctc_group_strategy: end
    brctc_risk_factor: 0.0
joint_net_conf: null
use_preprocessor: true
token_type: bpe
bpemodel: data/en_token_list/bpe_unigram500/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
frontend: default
frontend_conf:
    n_fft: 512
    win_length: 400
    hop_length: 160
    fs: 16k
specaug: specaug
specaug_conf:
    apply_time_warp: false
    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: 10
normalize: global_mvn
normalize_conf:
    stats_file: /scratch/bbjs/arora1/new_download_espnet_egs2/harpervalley/slu1_superb_onlyda/owsm_v3.1_ebf/exp/s2t_stats_raw_bpe50000/train/feats_stats.npz
model: espnet
model_conf:
    ctc_weight: 0.0
    lsm_weight: 0.1
    length_normalized_loss: false
    weighted_sum: true
    extract_feats_in_collect_stats: false
preencoder: null
preencoder_conf: {}
encoder: e_branchformer
encoder_conf:
    output_size: 1024
    attention_heads: 16
    attention_layer_type: selfattn
    pos_enc_layer_type: abs_pos
    rel_pos_type: latest
    cgmlp_linear_units: 4096
    cgmlp_conv_kernel: 31
    use_linear_after_conv: false
    gate_activation: identity
    num_blocks: 18
    dropout_rate: 0.2
    positional_dropout_rate: 0.2
    attention_dropout_rate: 0.2
    input_layer: conv2d
    layer_drop_rate: 0.0
    linear_units: 4096
    positionwise_layer_type: linear
    use_ffn: true
    macaron_ffn: true
    merge_conv_kernel: 31
prepostencoder: linear
prepostencoder_conf:
    input_size: 1024
    output_size: 80
postencoder: conformer_full
postencoder_conf:
    output_size: 256
    attention_heads: 4
    linear_units: 1024
    num_blocks: 12
    dropout_rate: 0.1
    positional_dropout_rate: 0.2
    attention_dropout_rate: 0.1
    input_layer: conv2d1
    normalize_before: true
    macaron_style: true
    rel_pos_type: latest
    pos_enc_layer_type: rel_pos
    selfattention_layer_type: rel_selfattn
    activation_type: swish
    use_cnn_module: true
    cnn_module_kernel: 31
deliberationencoder: null
deliberationencoder_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
postdecoder: null
postdecoder_conf: {}
required:
- output_dir
- token_list
version: '202310'
distributed: true

</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}
}