--- tags: - espnet - audio - automatic-speech-recognition language: te datasets: - microsoft_indian_languages_interspeech2018 license: cc-by-4.0 --- ## ESPnet2 model ### `` This model was trained by Chaitanya Narisetty using recipe in [espnet](https://github.com/espnet/espnet/). ### Demo: How to use in ESPnet2 ```bash cd espnet pip install -e . cd egs2/ms_indic_is18/asr1 ./run.sh --skip_data_prep false --skip_train true --download_model espnet/chai_microsoft_indian_langs_te ``` # RESULTS ## Environments - date: `Tue Mar 22 13:38:24 EDT 2022` - python version: `3.9.5 (default, Jun 4 2021, 12:28:51) [GCC 7.5.0]` - espnet version: `espnet 0.10.7a1` - pytorch version: `pytorch 1.8.1+cu111` - Git hash: `f91410f712d1287cd6809c5bf26b54c5a40fe314` - Commit date: `Mon Mar 14 22:32:17 2022 -0400` ## asr_train_asr_xlsr53_conformer_raw_te_bpe150_sp ### WER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |decode_transformer5_lm_lm_train_lm_rnn_te_bpe150_valid.loss.ave_asr_model_valid.acc.ave/test_te|3040|28413|78.0|19.5|2.5|2.4|24.4|80.1| |decode_transformer5_lm_lm_train_lm_rnn_te_bpe150_valid.loss.best_asr_model_valid.acc.ave/test_te|3040|28413|78.0|19.4|2.6|2.4|24.4|79.7| |decode_transformer5_lm_lm_train_lm_transformer_te_bpe150_valid.loss.ave_asr_model_valid.acc.ave/test_te|3040|28413|78.0|19.5|2.6|2.5|24.5|79.9| ### CER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |decode_transformer5_lm_lm_train_lm_rnn_te_bpe150_valid.loss.ave_asr_model_valid.acc.ave/test_te|3040|229419|95.6|2.2|2.2|1.6|6.1|80.1| |decode_transformer5_lm_lm_train_lm_rnn_te_bpe150_valid.loss.best_asr_model_valid.acc.ave/test_te|3040|229419|95.6|2.2|2.2|1.6|6.0|79.7| |decode_transformer5_lm_lm_train_lm_transformer_te_bpe150_valid.loss.ave_asr_model_valid.acc.ave/test_te|3040|229419|95.6|2.1|2.2|1.6|6.0|79.9| ### TER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |decode_transformer5_lm_lm_train_lm_rnn_te_bpe150_valid.loss.ave_asr_model_valid.acc.ave/test_te|3040|146657|92.7|4.7|2.6|1.6|8.9|80.1| |decode_transformer5_lm_lm_train_lm_rnn_te_bpe150_valid.loss.best_asr_model_valid.acc.ave/test_te|3040|146657|92.8|4.7|2.6|1.6|8.9|79.7| |decode_transformer5_lm_lm_train_lm_transformer_te_bpe150_valid.loss.ave_asr_model_valid.acc.ave/test_te|3040|146657|92.8|4.6|2.6|1.6|8.9|79.9| ## config
expand ``` config: conf/tuning/train_asr_xlsr53_conformer.yaml print_config: false log_level: INFO dry_run: false iterator_type: sequence output_dir: exp/asr_train_asr_xlsr53_conformer_raw_te_bpe150_sp 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: 50 patience: 15 val_scheduler_criterion: - valid - loss early_stopping_criterion: - valid - loss - min best_model_criterion: - - valid - acc - max keep_nbest_models: 5 nbest_averaging_interval: 0 grad_clip: 5 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_matplotlib: true 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: - frontend.upstream num_iters_per_epoch: null batch_size: 64 valid_batch_size: null batch_bins: 1000000 valid_batch_bins: null train_shape_file: - exp/asr_stats_raw_te_bpe150_sp_ssl/train/speech_shape - exp/asr_stats_raw_te_bpe150_sp_ssl/train/text_shape.bpe valid_shape_file: - exp/asr_stats_raw_te_bpe150_sp_ssl/valid/speech_shape - exp/asr_stats_raw_te_bpe150_sp_ssl/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_te_sp/wav.scp - speech - sound - - dump/raw/train_te_sp/text - text - text valid_data_path_and_name_and_type: - - dump/raw/dev_te/wav.scp - speech - sound - - dump/raw/dev_te/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: 0.0005 scheduler: warmuplr scheduler_conf: warmup_steps: 30000 token_list: - - - ా - ు - ి - ం - ే - వ - న - ల - ▁అ - క - ్ - ో - మ - ▁ - త - ర - ప - ీ - ▁మ - య - డ - ▁ప - ద - ని - గ - ▁వ - స - కు - ె - ర్ - ▁స - ▁క - ్య - న్న - ట - ▁చ - ▁త - ాల - ంట - ూ - శ - ంద - ార - ▁న - ారు - ▁ఉ - లు - ▁ఆ - ను - జ - రి - ▁ప్ర - ించ - ధ - ై - హ - ంది - ్ర - ▁ఇ - చ - రు - స్త - లో - ▁ద - డు - ▁ఎ - ▁వి - ల్ల - ణ - గా - ది - డి - న్నారు - దు - ిన - ▁ర - త్ - ొ - ▁గ - ంత - ంగా - ▁కా - బ - ▁జ - ష - ▁తెల - ులు - ▁ఏ - ట్ట - చ్చ - తి - నే - కి - ంలో - ▁అవును - ▁చెప్ప - భ - ▁ఈ - ప్ప - ▁ని - ▁రా - క్క - ▁బ - ట్ల - ▁భ - తో - ▁కూడా - ▁బా - ద్ద - ▁చేస - ▁లే - ాయి - ానికి - త్ర - ▁కొ - ఖ - ▁ఒక - ▁చాలా - క్ష - ళ - ▁చేస్త - ృ - థ - ఘ - ఫ - ఓ - ౌ - ఒ - ఐ - ఠ - ఢ - అ - ఉ - ఏ - ఈ - ౦ - ఇ - ః - ఋ - ఝ - ఔ - ఛ - ఞ - ఊ - ఎ - ఆ - ఙ - init: xavier_uniform input_size: null ctc_conf: dropout_rate: 0.0 ctc_type: builtin reduce: true ignore_nan_grad: true joint_net_conf: null model_conf: ctc_weight: 0.3 lsm_weight: 0.1 length_normalized_loss: false extract_feats_in_collect_stats: false use_preprocessor: true token_type: bpe bpemodel: data/te_token_list/bpe_unigram150/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: fused frontend_conf: frontends: - frontend_type: default n_fft: 512 win_length: 400 hop_length: 160 - frontend_type: s3prl frontend_conf: upstream: wav2vec2_xlsr download_dir: ./hub multilayer_feature: true align_method: linear_projection proj_dim: 200 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: 400 output_size: 100 encoder: conformer 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: 0.1 input_layer: conv2d normalize_before: true macaron_style: true pos_enc_layer_type: rel_pos selfattention_layer_type: rel_selfattn activation_type: swish use_cnn_module: true cnn_module_kernel: 15 postencoder: null postencoder_conf: {} decoder: transformer decoder_conf: input_layer: embed num_blocks: 6 linear_units: 2048 dropout_rate: 0.1 positional_dropout_rate: 0.1 self_attention_dropout_rate: 0.1 src_attention_dropout_rate: 0.1 required: - output_dir - token_list version: 0.10.7a1 distributed: false ```
### Citing ESPnet ```BibTex @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: ```bibtex @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} } ```