--- tags: - espnet - audio - automatic-speech-recognition language: en datasets: - an4 license: cc-by-4.0 --- ## ESPnet2 ASR model ### `Fhrozen/test_an4` This model was trained by Fhrozen using an4 recipe in [espnet](https://github.com/espnet/espnet/). ### Demo: How to use in ESPnet2 ```bash cd espnet git checkout b8df4c928e132acff78d196988bdb68a66987952 pip install -e . cd egs2/an4/asr1 ./run.sh --skip_data_prep false --skip_train true --download_model Fhrozen/test_an4 ``` # RESULTS ## Environments - date: `Wed Oct 20 00:00:46 JST 2021` - python version: `3.9.7 (default, Sep 16 2021, 13:09:58) [GCC 7.5.0]` - espnet version: `espnet 0.10.4a1` - pytorch version: `pytorch 1.9.0` - Git hash: `b8df4c928e132acff78d196988bdb68a66987952` - Commit date: `Tue Oct 19 07:48:11 2021 -0400` ## asr_train_raw_en_bpe30 ### WER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |inference_lm_lm_train_lm_en_bpe30_valid.loss.ave_asr_model_valid.acc.best/test|130|773|4.0|22.3|73.7|0.1|96.1|100.0| |inference_lm_lm_train_lm_en_bpe30_valid.loss.ave_asr_model_valid.acc.best/train_dev|100|591|2.7|21.8|75.5|0.0|97.3|100.0| ### CER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |inference_lm_lm_train_lm_en_bpe30_valid.loss.ave_asr_model_valid.acc.best/test|130|2565|17.2|16.4|66.4|1.0|83.8|100.0| |inference_lm_lm_train_lm_en_bpe30_valid.loss.ave_asr_model_valid.acc.best/train_dev|100|1915|15.5|16.4|68.1|0.9|85.5|100.0| ### TER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |inference_lm_lm_train_lm_en_bpe30_valid.loss.ave_asr_model_valid.acc.best/test|130|2695|21.1|15.6|63.3|0.9|79.9|100.0| |inference_lm_lm_train_lm_en_bpe30_valid.loss.ave_asr_model_valid.acc.best/train_dev|100|2015|19.4|15.6|65.0|0.9|81.5|100.0| ## ASR config
expand ``` config: null print_config: false log_level: INFO dry_run: false iterator_type: sequence output_dir: exp/asr_train_raw_en_bpe30 ngpu: 0 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: null 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: 40 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: - 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 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_bpe30/train/speech_shape - exp/asr_stats_raw_en_bpe30/train/text_shape.bpe valid_shape_file: - exp/asr_stats_raw_en_bpe30/valid/speech_shape - exp/asr_stats_raw_en_bpe30/valid/text_shape.bpe 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_nodev/wav.scp - speech - sound - - dump/raw/train_nodev/text - text - text valid_data_path_and_name_and_type: - - dump/raw/train_dev/wav.scp - speech - sound - - dump/raw/train_dev/text - text - text allow_variable_data_keys: false max_cache_size: 0.0 max_cache_fd: 32 valid_max_cache_size: null optim: adadelta optim_conf: {} scheduler: null scheduler_conf: {} token_list: - - - ▁ - T - E - O - R - Y - A - H - U - S - I - F - B - L - P - D - G - M - C - V - X - J - K - Z - W - N - Q - 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: bpe bpemodel: data/en_token_list/bpe_unigram30/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' frontend: default frontend_conf: fs: 16k specaug: null specaug_conf: {} normalize: global_mvn normalize_conf: stats_file: exp/asr_stats_raw_en_bpe30/train/feats_stats.npz preencoder: null preencoder_conf: {} encoder: rnn encoder_conf: {} postencoder: null postencoder_conf: {} decoder: rnn decoder_conf: {} required: - output_dir - token_list version: 0.10.4a1 distributed: false ```
## LM config
expand ``` config: conf/train_lm.yaml print_config: false log_level: INFO dry_run: false iterator_type: sequence output_dir: exp/lm_train_lm_en_bpe30 ngpu: 0 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: null 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: 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_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: 256 valid_batch_size: null batch_bins: 1000000 valid_batch_bins: null train_shape_file: - exp/lm_stats_en_bpe30/train/text_shape.bpe valid_shape_file: - exp/lm_stats_en_bpe30/valid/text_shape.bpe 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/lm_train.txt - text - text valid_data_path_and_name_and_type: - - dump/raw/train_dev/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: 0.1 scheduler: null scheduler_conf: {} token_list: - - - ▁ - T - E - O - R - Y - A - H - U - S - I - F - B - L - P - D - G - M - C - V - X - J - K - Z - W - N - Q - init: null model_conf: ignore_id: 0 use_preprocessor: true token_type: bpe bpemodel: data/en_token_list/bpe_unigram30/bpe.model non_linguistic_symbols: null cleaner: null g2p: null lm: seq_rnn lm_conf: unit: 650 nlayers: 2 required: - output_dir - token_list version: 0.10.4a1 distributed: false ```