--- tags: - espnet - audio - automatic-speech-recognition language: en3 datasets: - esc50 license: cc-by-4.0 --- ## ESPnet2 ASR model ### `shikhar7ssu/BEATs-ESC-FinetunedFold3` This model was trained by Shikhar Bharadwaj using esc50 recipe in [espnet](https://github.com/espnet/espnet/). ### Demo: How to use in ESPnet2 Follow the [ESPnet installation instructions](https://espnet.github.io/espnet/installation.html) if you haven't done that already. ```bash cd espnet git checkout ca9ca1ef8bc86753238ca7a0de05d87b8f57abb3 pip install -e . cd egs2/esc50/asr1 ./run.sh --skip_data_prep false --skip_train true --download_model shikhar7ssu/BEATs-ESC-FinetunedFold3 ``` # RESULTS ## Environments - date: `Sat Dec 14 18:40:28 EST 2024` - python version: `3.9.20 (main, Oct 3 2024, 07:27:41) [GCC 11.2.0]` - espnet version: `espnet 202412` - pytorch version: `pytorch 2.4.0` - Git hash: `cb80e61a15d6a13dc342ae5a413d2b870dd869c6` - Commit date: `Fri Dec 13 11:57:16 2024 -0500` ## /compute/babel-13-33/sbharad2/expdir/asr_fast.fold3/inference_ctc_weight0.0_maxlenratio-1_asr_model_valid.acc.best ### WER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |org/val3|400|400|94.8|5.3|0.0|0.0|5.3|5.3| ### CER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |org/val3|400|5520|99.4|0.5|0.0|0.0|0.6|5.3| ### TER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| ## ASR config
expand ``` config: conf/beats_classification.yaml print_config: false log_level: INFO drop_last_iter: false dry_run: false iterator_type: sequence valid_iterator_type: null output_dir: /compute/babel-13-33/sbharad2/expdir/asr_fast.fold3 ngpu: 1 seed: 0 num_workers: 2 num_att_plot: 0 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: true sharded_ddp: false use_deepspeed: false deepspeed_config: null cudnn_enabled: true cudnn_benchmark: false cudnn_deterministic: true use_tf32: false collect_stats: false write_collected_feats: false max_epoch: 1000 patience: null val_scheduler_criterion: - valid - loss early_stopping_criterion: - valid - loss - min best_model_criterion: - - valid - acc - max keep_nbest_models: 1 nbest_averaging_interval: 0 grad_clip: 1 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 create_graph_in_tensorboard: false use_wandb: true wandb_project: BEATs-ESC wandb_id: null wandb_entity: shikhar wandb_name: fast.fold3 wandb_model_log_interval: 0 detect_anomaly: false use_adapter: false adapter: lora save_strategy: all adapter_conf: {} pretrain_path: null init_param: [] ignore_init_mismatch: false freeze_param: [] num_iters_per_epoch: null batch_size: 128 valid_batch_size: null batch_bins: 1000000 valid_batch_bins: null category_sample_size: 10 train_shape_file: - /compute/babel-13-33/sbharad2/expdir/asr_stats_raw_3_word/train/speech_shape - /compute/babel-13-33/sbharad2/expdir/asr_stats_raw_3_word/train/text_shape.word valid_shape_file: - /compute/babel-13-33/sbharad2/expdir/asr_stats_raw_3_word/valid/speech_shape - /compute/babel-13-33/sbharad2/expdir/asr_stats_raw_3_word/valid/text_shape.word batch_type: folded valid_batch_type: null fold_length: - 100000 - 150 sort_in_batch: descending shuffle_within_batch: false sort_batch: descending multiple_iterator: false chunk_length: 500 chunk_shift_ratio: 0.5 num_cache_chunks: 1024 chunk_excluded_key_prefixes: [] chunk_default_fs: null chunk_max_abs_length: null chunk_discard_short_samples: true train_data_path_and_name_and_type: - - /compute/babel-13-33/sbharad2/dumpdir/raw/train3/wav.scp - speech - sound - - /compute/babel-13-33/sbharad2/dumpdir/raw/train3/text - text - text valid_data_path_and_name_and_type: - - /compute/babel-13-33/sbharad2/dumpdir/raw/val3/wav.scp - speech - sound - - /compute/babel-13-33/sbharad2/dumpdir/raw/val3/text - text - text multi_task_dataset: false allow_variable_data_keys: false max_cache_size: 0.0 max_cache_fd: 32 allow_multi_rates: false valid_max_cache_size: null exclude_weight_decay: false exclude_weight_decay_conf: {} optim: adamw optim_conf: lr: 0.0001 weight_decay: 0.01 betas: - 0.9 - 0.98 scheduler: cosineannealingwarmuprestarts scheduler_conf: first_cycle_steps: 6000 warmup_steps: 300 max_lr: 0.0001 min_lr: 5.0e-06 token_list: - - - audio_class:0 - audio_class:14 - audio_class:36 - audio_class:19 - audio_class:30 - audio_class:34 - audio_class:9 - audio_class:22 - audio_class:48 - audio_class:41 - audio_class:47 - audio_class:31 - audio_class:17 - audio_class:45 - audio_class:8 - audio_class:15 - audio_class:46 - audio_class:37 - audio_class:32 - audio_class:16 - audio_class:25 - audio_class:4 - audio_class:3 - audio_class:27 - audio_class:43 - audio_class:12 - audio_class:40 - audio_class:29 - audio_class:10 - audio_class:7 - audio_class:26 - audio_class:6 - audio_class:44 - audio_class:23 - audio_class:20 - audio_class:49 - audio_class:24 - audio_class:39 - audio_class:28 - audio_class:18 - audio_class:2 - audio_class:35 - audio_class:38 - audio_class:21 - audio_class:1 - audio_class:11 - audio_class:42 - audio_class:5 - audio_class:33 - audio_class:13 - init: xavier_normal input_size: 1 ctc_conf: dropout_rate: 0.0 ctc_type: builtin reduce: true ignore_nan_grad: null zero_infinity: true brctc_risk_strategy: exp brctc_group_strategy: end brctc_risk_factor: 0.0 joint_net_conf: null use_preprocessor: true use_lang_prompt: false use_nlp_prompt: false 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' short_noise_thres: 0.5 aux_ctc_tasks: [] frontend: null frontend_conf: {} specaug: null specaug_conf: {} normalize: null normalize_conf: {} model: espnet model_conf: ctc_weight: 0.0 lsm_weight: 0.1 length_normalized_loss: true preencoder: null preencoder_conf: {} encoder: beats encoder_conf: beats_ckpt_path: /compute/babel-13-33/sbharad2/models/BEATs/BEATs_iter3.pt fbank_mean: 11.72215 fbank_std: 10.60431 beats_config: layer_wise_gradient_decay_ratio: 0.2 encoder_layerdrop: 0.1 dropout: 0.0 specaug_config: apply_time_warp: true apply_freq_mask: false freq_mask_width_range: - 0 - 32 num_freq_mask: 1 apply_time_mask: true time_mask_width_ratio_range: - 0 - 0.06 num_time_mask: 1 roll_augment: true roll_interval: 16000 use_weighted_representation: false postencoder: null postencoder_conf: {} decoder: linear_decoder decoder_conf: pooling: mean dropout: 0.1 preprocessor: default preprocessor_conf: {} required: - output_dir - token_list version: '202412' 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} } ```