--- tags: - espnet - audio - self-supervised-learning language: en datasets: - librispeech license: cc-by-4.0 --- ## ESPnet2 SSL model ### `espnet/hubert_dummy` This model was trained by chen26 using librispeech 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 pip install -e . cd egs2/librispeech/ssl1 ./run.sh --skip_data_prep false --skip_train true --download_model espnet/hubert_dummy ``` ## SSL config
expand ``` config: conf/tuning/train_hubert_dummy.yaml print_config: false log_level: INFO drop_last_iter: false dry_run: false iterator_type: sequence valid_iterator_type: null output_dir: exp/ssl_train_hubert_dummy_raw ngpu: 1 seed: 0 num_workers: 4 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: false sharded_ddp: false use_deepspeed: true deepspeed_config: conf/deepspeed.json gradient_as_bucket_view: true ddp_comm_hook: null cudnn_enabled: true cudnn_benchmark: false cudnn_deterministic: true use_tf32: false collect_stats: false write_collected_feats: false max_epoch: 1 patience: null val_scheduler_criterion: - valid - loss early_stopping_criterion: - valid - loss - min best_model_criterion: - - valid - total_count - max keep_nbest_models: 5 nbest_averaging_interval: 0 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_matplotlib: true use_tensorboard: true create_graph_in_tensorboard: false use_wandb: false wandb_project: null wandb_id: null wandb_entity: null wandb_name: null wandb_model_log_interval: -1 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: 10 batch_size: 20 valid_batch_size: null batch_bins: 16000 valid_batch_bins: null category_sample_size: 10 train_shape_file: - exp/ssl_stats_raw/train/speech_shape valid_shape_file: - exp/ssl_stats_raw/valid/speech_shape batch_type: numel valid_batch_type: null fold_length: - 80000 - 400 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: - - dump/raw/train_960/wav.scp - speech - sound - - dump/raw/train_960/text - text - text valid_data_path_and_name_and_type: - - dump/raw/dev/wav.scp - speech - sound - - dump/raw/dev/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: adadelta optim_conf: {} scheduler: null scheduler_conf: {} token_list: - '30' - '4' - '72' - '305' - '275' - '24' - '369' - '125' - '202' - '368' - '270' - '296' - '68' - '188' - '418' - '223' - '8' - '338' - '437' - '14' - '299' - '469' - '415' - '11' - '41' - '227' - '44' - '35' - '179' - '449' - '23' - '10' - '416' - '291' - '100' - '74' - '327' - '107' - '321' - '208' - '76' - '267' - '130' - '173' - '96' - '162' - '456' - '84' - '98' - '217' - '48' - '482' - '127' - '110' - '366' - '336' - '387' - '105' - '373' - '139' - '61' - '370' - '464' - '397' - '281' - '151' - '154' - '155' - '203' - '440' - '119' - '71' - '320' - '93' - '20' - '138' - '78' - '216' - '104' - '205' - '38' - '382' - '238' - '474' - '225' - '465' - '309' - '17' - '285' - '90' - '375' - '356' - '256' - '392' - '311' - '398' - '9' - '264' - '341' - '168' - '339' - '40' - '344' - '422' - '63' - '396' - '51' - '184' - '441' - '346' - '252' - '206' - '322' - '444' - '198' - '66' - '269' - '145' - '69' - '244' - '463' - '37' - '172' - '271' - '313' - '279' - '106' - '377' - '158' - '5' - '445' - '455' - '134' - '287' - '7' - '297' - '420' - '13' - '31' - '484' - '91' - '34' - '488' - '468' - '21' - '193' - '288' - '159' - '247' - '476' - '25' - '265' - '115' - '50' - '394' - '197' - '116' - '57' - '182' - '378' - '135' - '89' - '167' - '19' - '148' - '425' - '103' - '95' - '454' - '376' - '178' - '79' - '424' - '261' - '36' - '426' - '152' - '102' - '292' - '258' - '60' - '328' - '280' - '273' - '111' - '240' - '213' - '483' - '300' - '363' - '174' - '317' - '419' - '439' - '42' - '118' - '222' - '15' - '276' - '277' - '166' - '304' - '114' - '329' - '395' - '413' - '435' - '33' - '266' - '133' - '210' - '408' - '330' - '315' - '251' - '6' - '357' - '171' - '56' - '1' - '59' - '359' - '28' - '215' - '97' - '274' - '170' - '49' - '81' - '108' - '282' - '85' - '200' - '80' - '243' - '364' - '113' - '176' - '433' - '77' - '335' - '231' - '462' - '62' - '286' - '67' - '191' - '228' - '16' - '22' - '122' - '235' - '331' - '137' - '289' - '92' - '157' - '417' - '319' - '2' - '101' - '129' - '169' - '26' - '165' - '143' - '229' - '220' - '324' - '393' - '272' - '43' - '367' - '204' - '410' - '278' - '73' - '65' - '428' - '411' - '380' - '99' - '83' - '412' - '307' - '306' - '201' - '361' - '232' - '290' - '109' - '140' - '438' - '64' - '447' - '374' - '301' - '249' - '186' - '234' - '121' - '239' - '255' - '82' - '384' - '160' - '494' - '351' - '283' - '32' - '54' - '52' - '187' - '337' - '112' - '260' - '132' - '47' - '457' - '211' - '490' - '430' - '423' - '175' - '142' - '499' - '407' - '303' - '12' - '403' - '209' - '233' - '262' - '146' - '436' - '219' - '316' - '123' - '460' - '39' - '58' - '333' - '475' - '70' - '218' - '199' - '295' - '389' - '345' - '156' - '383' - '390' - '192' - '343' - '150' - '318' - '196' - '94' - '194' - '27' - '459' - '257' - '371' - '498' - '485' - '190' - '402' - '163' - '491' - '0' - '241' - '467' - '149' - '18' - '429' - '421' - '189' - '365' - '3' - '75' - '141' - '259' - '120' - '372' - '405' - '354' - '446' - '340' - '406' - '353' - '53' - '334' - '427' - '432' - '442' - '131' - '88' - '470' - '473' - '254' - '349' - '214' - '153' - '342' - '212' - '434' - '46' - '86' - '350' - '284' - '308' - '323' - '381' - '161' - '391' - '248' - '180' - '230' - '452' - '325' - '246' - '224' - '347' - '195' - '128' - '55' - '314' - '126' - '147' - '481' - '185' - '358' - '478' - '400' - '495' - '388' - '177' - '181' - '466' - '362' - '268' - '326' - '144' - '493' - '489' - '450' - '399' - '443' - '253' - '236' - '117' - '448' - '312' - '379' - '492' - '496' - '87' - '332' - '298' - '497' - '221' - '480' - '226' - '302' - '348' - '136' - '451' - '479' - '183' - '45' - '404' - '263' - '477' - '355' - '29' - '414' - '237' - '409' - '385' - '461' - '386' - '124' - '401' - '352' - '293' - '471' - '458' - '472' - '486' - '164' - '453' - '310' - '207' - '487' - '294' - '360' - '245' - '242' - '431' - '250' - - init: null collate_fn_conf: label_downsampling: 1 pad: false rand_crop: true input_size: null num_classes: null use_preprocessor: true 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' window_size: null window_shift: null loss: - name: hubert conf: num_classes: 500 final_dim: 2 util: - name: mask conf: {} frontend: wav2vec_cnn frontend_conf: norm_mode: group_norm conv_mode: standard bias: false normalize_audio: false shapes: - - 2 - 1 - 10 fs: 16k specaug: null specaug_conf: {} normalize: null normalize_conf: {} preencoder: linear preencoder_conf: output_size: 16 encoder: transformer encoder_conf: output_size: 16 attention_heads: 1 linear_units: 4 num_blocks: 2 dropout_rate: 0.1 positional_dropout_rate: 0.0 attention_dropout_rate: 0.1 input_layer: wav2vec normalize_before: false pos_enc_layer_type: conv model: espnet model_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} } ```