Published November 9, 2021 | Version v1
Other Open

ESPnet2 pretrained model, Siddhant/fsc_challenge_asr_train_asr_hubert_transformer_adam_specaug_raw_en_word_valid.acc.ave_5best, fs=16k, lang=en

Authors/Creators

Description

This model was trained by Siddhant using fsc_challenge 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 97b9dad4dbca71702cb7928a126ec45d96414a3f
pip install -e .
cd egs2/fsc_challenge/asr1
./run.sh --skip_data_prep false --skip_train true --download_model Siddhant/fsc_challenge_asr_train_asr_hubert_transformer_adam_specaug_raw_en_word_valid.acc.ave_5best
  • Results 
    
# RESULTS

## Environments
- date: `Sun Oct  3 22:25:25 EDT 2021`
- python version: `3.8.11 (default, Aug  3 2021, 15:09:35)  [GCC 7.5.0]`
- espnet version: `espnet 0.10.3a3`
- pytorch version: `pytorch 1.9.0+cu102`
- Git hash: `a1a55e1eef2a74d2b8580d8071ce5229e7fa654c`
  - Commit date: `Mon Nov 8 23:56:06 2021 -0500`


## Using Transformer based encoder-decoder with finetuned Hubert pre encoder and decoding sentence with spectral augmentation and predicting transcript along with intent
- ASR config: [conf/tuning/train_asr.yaml](conf/tuning/train_asr_hubert_transformer_adam_specaug_finetune.yaml)
- token_type: word
- keep_nbest_models: 5

|dataset|Snt|Intent Classification (%)|
|---|---|---|
|inference_asr_model_valid.acc.ave_5best/spk_test|3366|97.5|
|inference_asr_model_valid.acc.ave_5best/utt_test|3970|78.5|
|inference_asr_model_valid.acc.ave_5best/valid|2624|98.4|

###ASR results

#### WER

|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|---|---|---|---|---|---|---|---|---|
|inference_asr_model_valid.acc.ave_5best/spk_test|3349|14588|98.7|0.9|0.4|0.6|1.9|4.7|
|inference_asr_model_valid.acc.ave_5best/utt_test|4201|18330|87.1|10.6|2.3|3.8|16.7|44.6|
|inference_asr_model_valid.acc.ave_5best/valid|2597|1185|98.9|0.6|0.5|0.3|1.3|2.9 |
  • ASR config 
    config: conf/tuning/train_asr_hubert_transformer_adam_specaug.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: sequence
output_dir: exp/asr_train_asr_hubert_transformer_adam_specaug_raw_en_word
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: 80
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - train
    - loss
    - min
-   - valid
    - loss
    - min
-   - train
    - acc
    - max
-   - valid
    - acc
    - max
keep_nbest_models: 5
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
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: 20
valid_batch_size: null
batch_bins: 1000000
valid_batch_bins: null
train_shape_file:
- exp/asr_stats_raw_en_word/train/speech_shape
- exp/asr_stats_raw_en_word/train/text_shape.word
valid_shape_file:
- exp/asr_stats_raw_en_word/valid/speech_shape
- exp/asr_stats_raw_en_word/valid/text_shape.word
batch_type: folded
valid_batch_type: null
fold_length:
- 80000
- 150
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/wav.scp
    - speech
    - sound
-   - dump/raw/train/text
    - text
    - text
valid_data_path_and_name_and_type:
-   - dump/raw/valid/wav.scp
    - speech
    - sound
-   - dump/raw/valid/text
    - 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: 0.0002
scheduler: warmuplr
scheduler_conf:
    warmup_steps: 25000
token_list:
- 
- 
- ▁the
- ▁
- e
- ▁turn
- ▁in
- s
- ▁lights
- o
- ▁m
- c
- i
- ▁heat
- a
- t
- hroom
- ▁up
- ▁s
- ▁on
- ▁down
- n
- ▁temperature
- crease
- p
- ▁t
- u
- ▁b
- ▁switch
- w
- h
- d
- ou
- ▁kitchen
- ▁volume
- ▁off
- ing
- y
- increase_volume_none
- ▁bedroom
- ▁langu
- age
- as
- decrease_volume_none
- ▁l
- r
- er
- at
- ▁d
- l
- decrease_heat_washroom
- increase_heat_washroom
- k
- an
- g
- increase_heat_none
- oo
- decrease_heat_none
- ge
- change_language_none_none
- activate_lights_washroom
- activate_lights_kitchen
- ow
- in
- activate_music_none
- mp
- deactivate_music_none
- increase_heat_bedroom
- increase_heat_kitchen
- decrease_heat_kitchen
- it
- activate_lights_bedroom
- deactivate_lights_bedroom
- f
- re
- decrease_heat_bedroom
- ed
- deactivate_lights_kitchen
- bring_newspaper_none
- bring_shoes_none
- bring_socks_none
- activate_lights_none
- deactivate_lights_none
- q
- deactivate_lights_washroom
- change_language_Chinese_none
- bring_juice_none
- j
- m
- deactivate_lamp_none
- activate_lamp_none
- change_language_Korean_none
- ▁k
- me
- change_language_German_none
- ▁o
- change_language_English_none
- ▁he
- ase
- ff
- ume
- ▁v
- x
- ▁u
- v
- 
init: null
input_size: null
ctc_conf:
    dropout_rate: 0.0
    ctc_type: builtin
    reduce: true
    ignore_nan_grad: true
model_conf:
    ctc_weight: 0.5
    ignore_id: -1
    lsm_weight: 0.0
    length_normalized_loss: false
    report_cer: true
    report_wer: true
    sym_space: 
    sym_blank: 
    extract_feats_in_collect_stats: true
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'
frontend: s3prl
frontend_conf:
    frontend_conf:
        upstream: hubert_large_ll60k
    download_dir: ./hub
    multilayer_feature: true
    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
    - 30
    num_freq_mask: 2
    apply_time_mask: true
    time_mask_width_range:
    - 0
    - 40
    num_time_mask: 2
normalize: utterance_mvn
normalize_conf: {}
preencoder: linear
preencoder_conf:
    input_size: 1024
    output_size: 80
encoder: transformer
encoder_conf:
    output_size: 256
    attention_heads: 4
    linear_units: 2048
    num_blocks: 12
    dropout_rate: 0.1
    positional_dropout_rate: 0.1
    attention_dropout_rate: 0.0
    input_layer: conv2d
    normalize_before: true
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: 0.0
    src_attention_dropout_rate: 0.0
required:
- output_dir
- token_list
version: 0.10.3a3
distributed: false
  • LM config 
    NONE

Files

asr_train_asr_hubert_transformer_adam_specaug_raw_en_word_valid.acc.ave_5best.zip

Files (1.4 GB)

Additional details

Related works

Is supplement to
https://github.com/espnet/espnet (URL)