ESPnet2 pretrained model, kamo-naoyuki/wsj_asr_train_asr_transformer_raw_char_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 5b33951b08555889a1e39ca32dee26a63304bdd5
pip install -e .
cd egs2/wsj/asr1
./run.sh --skip_data_prep false --skip_train true --download_model kamo-naoyuki/wsj_asr_train_asr_transformer_raw_char_valid.acc.ave

# RESULTS
## Environments
- date: `Thu Aug 27 09:50:23 JST 2020`
- python version: `3.7.3 (default, Mar 27 2019, 22:11:17)  [GCC 7.3.0]`
- espnet version: `espnet 0.9.0`
- pytorch version: `pytorch 1.6.0`
- Git hash: `4b040785069f114fe046ffa9acdd8698fdeb7f21`
  - Commit date: `Tue Aug 25 17:46:25 2020 +0900`

## asr_train_asr_transformer_raw_char
### WER

|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|---|---|---|---|---|---|---|---|---|
|inference_lm_lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4_valid.loss.ave_asr_model_valid.acc.ave/test_dev93|503|8234|93.3|5.9|0.8|0.8|7.5|58.1|
|inference_lm_lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4_valid.loss.ave_asr_model_valid.acc.ave/test_eval92|333|5643|95.8|4.0|0.3|0.7|5.0|44.7|
|inference_lm_lm_train_lm_char_optimadam_optim_conflr0.0005_lm_confnlayers4_keep_nbest_models235810_batch_size512_valid.loss.ave_asr_model_valid.acc.ave/test_dev93|503|8234|93.2|5.9|0.8|0.9|7.6|59.0|
|inference_lm_lm_train_lm_char_optimadam_optim_conflr0.0005_lm_confnlayers4_keep_nbest_models235810_batch_size512_valid.loss.ave_asr_model_valid.acc.ave/test_eval92|333|5643|95.8|4.0|0.2|0.7|4.9|45.6|
|inference_lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4_valid.loss.best_asr_model_valid.acc.best/test_dev93|503|8234|92.3|6.8|1.0|1.0|8.8|62.2|
|inference_lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4_valid.loss.best_asr_model_valid.acc.best/test_eval92|333|5643|94.8|4.8|0.5|0.9|6.1|48.6|
|inference_lm_train_lm_char_valid.loss.best_asr_model_valid.acc.best/test_dev93|503|8234|80.9|17.4|1.7|2.6|21.7|90.7|
|inference_lm_train_lm_char_valid.loss.best_asr_model_valid.acc.best/test_eval92|333|5643|84.3|14.6|1.1|2.4|18.1|86.8|

### CER

|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|---|---|---|---|---|---|---|---|---|
|inference_lm_lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4_valid.loss.ave_asr_model_valid.acc.ave/test_dev93|503|48634|97.6|1.1|1.3|0.6|3.0|62.8|
|inference_lm_lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4_valid.loss.ave_asr_model_valid.acc.ave/test_eval92|333|33341|98.5|0.7|0.8|0.5|2.0|53.2|
|inference_lm_lm_train_lm_char_optimadam_optim_conflr0.0005_lm_confnlayers4_keep_nbest_models235810_batch_size512_valid.loss.ave_asr_model_valid.acc.ave/test_dev93|503|48634|97.5|1.2|1.3|0.5|3.1|63.6|
|inference_lm_lm_train_lm_char_optimadam_optim_conflr0.0005_lm_confnlayers4_keep_nbest_models235810_batch_size512_valid.loss.ave_asr_model_valid.acc.ave/test_eval92|333|33341|98.5|0.7|0.8|0.5|2.0|53.2|
|inference_lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4_valid.loss.best_asr_model_valid.acc.best/test_dev93|503|48634|97.0|1.4|1.6|0.7|3.6|67.2|
|inference_lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4_valid.loss.best_asr_model_valid.acc.best/test_eval92|333|33341|98.0|0.9|1.0|0.5|2.5|56.8|
|inference_lm_train_lm_char_valid.loss.best_asr_model_valid.acc.best/test_dev93|503|48634|94.4|2.8|2.9|1.5|7.1|91.3|
|inference_lm_train_lm_char_valid.loss.best_asr_model_valid.acc.best/test_eval92|333|33341|95.6|2.2|2.1|1.3|5.7|88.9|

config: conf/train_asr_transformer.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: sequence
output_dir: exp/asr_train_asr_transformer_raw_char
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
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 100
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_noise: false
accum_grad: 8
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: null
pretrain_path: []
pretrain_key: []
num_iters_per_epoch: null
batch_size: 32
valid_batch_size: null
batch_bins: 1000000
valid_batch_bins: null
train_shape_file:
- exp/asr_stats_raw/train/speech_shape
- exp/asr_stats_raw/train/text_shape.char
valid_shape_file:
- exp/asr_stats_raw/valid/speech_shape
- exp/asr_stats_raw/valid/text_shape.char
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_si284/wav.scp
    - speech
    - sound
-   - dump/raw/train_si284/text
    - text
    - text
valid_data_path_and_name_and_type:
-   - dump/raw/test_dev93/wav.scp
    - speech
    - sound
-   - dump/raw/test_dev93/text
    - text
    - text
allow_variable_data_keys: false
max_cache_size: 0.0
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
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: null
specaug_conf: {}
normalize: global_mvn
normalize_conf:
    stats_file: exp/asr_stats_raw/train/feats_stats.npz
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: 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
distributed: false

config: conf/train_lm.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: sequence
output_dir: exp/lm_train_lm_char_optimadam_batch_size512_optim_conflr0.0005_lm_confnlayers4
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
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 40
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - valid
    - loss
    - min
keep_nbest_models: 1
grad_clip: 5.0
grad_noise: false
accum_grad: 1
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: null
pretrain_path: []
pretrain_key: []
num_iters_per_epoch: null
batch_size: 512
valid_batch_size: null
batch_bins: 1000000
valid_batch_bins: null
train_shape_file:
- exp/lm_stats/train/text_shape.char
valid_shape_file:
- exp/lm_stats/valid/text_shape.char
batch_type: folded
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/srctexts
    - text
    - text
valid_data_path_and_name_and_type:
-   - dump/raw/test_dev93/text
    - text
    - text
allow_variable_data_keys: false
max_cache_size: 0.0
valid_max_cache_size: null
optim: adam
optim_conf:
    lr: 0.0005
scheduler: null
scheduler_conf: {}
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: seq_rnn
lm_conf:
    unit: 650
    nlayers: 4
required:
- output_dir
- token_list
distributed: false