ESPnet2 pretrained model, kamo-naoyuki/chime4_asr_train_asr_transformer3_raw_en_char_sp_valid.acc.ave, fs=16k, lang=en

See https://github.com/espnet/espnet_model_zoo

git clone https://github.com/espnet/espnet
cd espnet
git checkout d5ddd5e601f064eea0c1f96acaeefa324fc7d392
pip install -e .
cd egs2/chime4/asr1
./run.sh --skip_data_prep false --skip_train true --download_model kamo-naoyuki/chime4_asr_train_asr_transformer3_raw_en_char_sp_valid.acc.ave

# RESULTS
## Environments
- date: `Wed Dec 30 06:42:56 JST 2020`
- python version: `3.8.5 (default, Sep  4 2020, 07:30:14)  [GCC 7.3.0]`
- espnet version: `espnet 0.9.6`
- pytorch version: `pytorch 1.4.0`
- Git hash: `d5ddd5e601f064eea0c1f96acaeefa324fc7d392`
  - Commit date: `Fri Dec 25 07:31:34 2020 +0000`

## asr_train_asr_transformer3_raw_en_char_sp
### WER

|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|---|---|---|---|---|---|---|---|---|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_real_beamformit_2mics|1640|27119|95.2|3.7|1.1|0.5|5.3|42.4|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_real_beamformit_5mics|1640|27119|96.2|2.9|0.9|0.4|4.3|37.9|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_real_isolated_1ch_track|1640|27119|94.3|4.4|1.3|0.7|6.3|46.6|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_simu_beamformit_2mics|1640|27120|93.3|5.1|1.6|0.8|7.5|52.0|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_simu_beamformit_5mics|1640|27120|94.9|3.8|1.3|0.6|5.7|45.9|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_simu_isolated_1ch_track|1640|27120|92.2|6.0|1.7|0.9|8.7|52.4|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_real_beamformit_2mics|1320|21409|91.4|6.9|1.7|1.0|9.6|52.8|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_real_beamformit_5mics|1320|21409|93.0|5.6|1.4|0.9|7.9|48.9|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_real_isolated_1ch_track|1320|21409|89.5|8.4|2.1|1.3|11.8|59.1|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_simu_beamformit_2mics|1320|21416|89.3|8.0|2.6|1.4|12.1|61.0|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_simu_beamformit_5mics|1320|21416|91.4|6.4|2.2|1.2|9.8|54.5|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_simu_isolated_1ch_track|1320|21416|87.5|9.6|2.8|1.6|14.1|60.8|

### CER

|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|---|---|---|---|---|---|---|---|---|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_real_beamformit_2mics|1640|160390|97.8|1.0|1.2|0.5|2.7|42.4|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_real_beamformit_5mics|1640|160390|98.3|0.8|0.9|0.4|2.1|37.9|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_real_isolated_1ch_track|1640|160390|97.4|1.3|1.3|0.7|3.3|46.6|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_simu_beamformit_2mics|1640|160400|96.7|1.5|1.8|0.8|4.1|52.0|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_simu_beamformit_5mics|1640|160400|97.6|1.1|1.3|0.6|3.0|45.9|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/dt05_simu_isolated_1ch_track|1640|160400|96.2|1.9|1.9|1.0|4.8|52.4|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_real_beamformit_2mics|1320|126796|95.9|2.1|2.0|1.0|5.1|52.8|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_real_beamformit_5mics|1320|126796|96.8|1.6|1.6|0.8|4.1|48.9|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_real_isolated_1ch_track|1320|126796|94.7|2.7|2.5|1.3|6.6|59.1|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_simu_beamformit_2mics|1320|126812|94.2|2.7|3.1|1.5|7.2|61.0|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_simu_beamformit_5mics|1320|126812|95.6|2.0|2.4|1.2|5.6|54.5|
|decode_asr_rnn_lm_lm_train_lm_transformer_en_char_valid.loss.ave_asr_model_valid.acc.ave/et05_simu_isolated_1ch_track|1320|126812|93.2|3.4|3.4|1.8|8.6|60.8|

config: conf/tuning/train_asr_transformer3.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: sequence
output_dir: exp/asr_train_asr_transformer3_raw_en_char_sp
ngpu: 1
seed: 0
num_workers: 1
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: 56935
dist_launcher: null
multiprocessing_distributed: true
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 50
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - valid
    - acc
    - max
keep_nbest_models: 10
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
unused_parameters: false
use_tensorboard: true
use_wandb: false
wandb_project: null
wandb_id: null
pretrain_path: null
init_param: []
freeze_param: []
num_iters_per_epoch: null
batch_size: 20
valid_batch_size: null
batch_bins: 16000000
valid_batch_bins: null
train_shape_file:
- exp/asr_stats_raw_en_char_sp/train/speech_shape
- exp/asr_stats_raw_en_char_sp/train/text_shape.char
valid_shape_file:
- exp/asr_stats_raw_en_char_sp/valid/speech_shape
- exp/asr_stats_raw_en_char_sp/valid/text_shape.char
batch_type: numel
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/tr05_multi_noisy_si284_sp/wav.scp
    - speech
    - sound
-   - dump/raw/tr05_multi_noisy_si284_sp/text
    - text
    - text
valid_data_path_and_name_and_type:
-   - dump/raw/dt05_multi_isolated_1ch_track/wav.scp
    - speech
    - sound
-   - dump/raw/dt05_multi_isolated_1ch_track/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: 0.005
scheduler: warmuplr
scheduler_conf:
    warmup_steps: 30000
token_list:
- 
- 
- 
- E
- T
- A
- N
- I
- O
- S
- R
- H
- L
- D
- C
- U
- M
- P
- F
- G
- Y
- W
- B
- V
- K
- .
- X
- ''''
- J
- Q
- Z
- ','
- '-'
- '"'
- 
- '*'
- ':'
- (
- )
- '?'
- '&'
- ;
- '!'
- /
- '{'
- '}'
- '1'
- '2'
- '0'
- $
- '8'
- '9'
- '6'
- '3'
- '5'
- '7'
- '4'
- '~'
- '`'
- _
- <*IN*>
- <*MR.*>
- \
- ^
- 
init: xavier_uniform
input_size: null
ctc_conf:
    dropout_rate: 0.0
    ctc_type: builtin
    reduce: true
    ignore_nan_grad: false
model_conf:
    ctc_weight: 0.3
    lsm_weight: 0.1
    length_normalized_loss: false
use_preprocessor: true
token_type: char
bpemodel: null
non_linguistic_symbols: data/nlsyms.txt
cleaner: null
g2p: null
frontend: default
frontend_conf:
    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: global_mvn
normalize_conf:
    stats_file: exp/asr_stats_raw_en_char_sp/train/feats_stats.npz
preencoder: null
preencoder_conf: {}
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
decoder: transformer
decoder_conf:
    attention_heads: 4
    linear_units: 2028
    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
distributed: true

config: conf/tuning/train_lm_transformer.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: sequence
output_dir: exp/lm_train_lm_transformer_en_char
ngpu: 1
seed: 0
num_workers: 1
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: 50129
dist_launcher: null
multiprocessing_distributed: true
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
    - loss
    - min
keep_nbest_models: 10
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: false
log_interval: null
unused_parameters: false
use_tensorboard: true
use_wandb: false
wandb_project: null
wandb_id: null
pretrain_path: null
init_param: []
freeze_param: []
num_iters_per_epoch: null
batch_size: 20
valid_batch_size: null
batch_bins: 350000
valid_batch_bins: null
train_shape_file:
- exp/lm_stats_en_char/train/text_shape.char
valid_shape_file:
- exp/lm_stats_en_char/valid/text_shape.char
batch_type: numel
valid_batch_type: null
fold_length:
- 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/lm_train.txt
    - text
    - text
valid_data_path_and_name_and_type:
-   - dump/raw/dt05_multi_isolated_1ch_track/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: 0.001
scheduler: warmuplr
scheduler_conf:
    warmup_steps: 25000
token_list:
- 
- 
- 
- E
- T
- A
- N
- I
- O
- S
- R
- H
- L
- D
- C
- U
- M
- P
- F
- G
- Y
- W
- B
- V
- K
- .
- X
- ''''
- J
- Q
- Z
- ','
- '-'
- '"'
- 
- '*'
- ':'
- (
- )
- '?'
- '&'
- ;
- '!'
- /
- '{'
- '}'
- '1'
- '2'
- '0'
- $
- '8'
- '9'
- '6'
- '3'
- '5'
- '7'
- '4'
- '~'
- '`'
- _
- <*IN*>
- <*MR.*>
- \
- ^
- 
init: null
model_conf:
    ignore_id: 0
use_preprocessor: true
token_type: char
bpemodel: null
non_linguistic_symbols: data/nlsyms.txt
cleaner: null
g2p: null
lm: transformer
lm_conf:
    pos_enc: null
    embed_unit: 128
    att_unit: 512
    head: 8
    unit: 2048
    layer: 16
    dropout_rate: 0.1
required:
- output_dir
- token_list
distributed: true