--- tags: - espnet - audio - automatic-speech-recognition language: - en datasets: - jibo_kids license: cc-by-4.0 --- ## ESPnet2 ASR model ### `balaji1312/jibo_kids_wavlm_aed_transformer` This model was trained by using 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/jibo_kids/asr1 ./run.sh --skip_data_prep false --skip_train true --download_model balaji1312/jibo_kids_wavlm_aed_transformer ``` # RESULTS ## Environments - date: `Thu Jan 30 06:18:01 EST 2025` - python version: `3.9.19 (main, May 6 2024, 19:43:03) [GCC 11.2.0]` - espnet version: `espnet 202402` - pytorch version: `pytorch 2.4.0` - Git hash: `c46aa9a594ff83d52cbf61d84c5650325d1ce527` - Commit date: `Sun Oct 13 14:39:31 2024 -0400` ## exp/asr_train_asr_wavlm_transformer_raw_en_bpe1024 ### WER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |decode_asr_asr_model_valid.acc.best/test|1044|3686|56.1|31.4|12.5|8.1|52.0|62.3| ### CER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |decode_asr_asr_model_valid.acc.best/test|1044|16215|75.4|8.1|16.6|9.4|34.1|62.3| ### TER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |decode_asr_asr_model_valid.acc.best/test|1044|5220|64.5|18.0|17.5|10.4|45.9|62.3| ## exp/asr_train_asr_wavlm_transformer_raw_en_bpe1024/decode_asr_asr_model_valid.acc.best ### WER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |org/dev|853|2372|59.8|31.2|8.9|7.2|47.3|64.0| ### CER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |org/dev|853|9855|78.3|7.3|14.3|8.4|30.1|64.0| ### TER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |org/dev|853|3590|68.2|16.2|15.6|6.4|38.3|64.0| ## ASR config
expand ``` config: conf/tuning/train_asr_wavlm_transformer.yaml print_config: false log_level: INFO drop_last_iter: false dry_run: false iterator_type: sequence valid_iterator_type: null output_dir: exp/asr_train_asr_wavlm_transformer_raw_en_bpe1024 ngpu: 1 seed: 2022 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: false deepspeed_config: null cudnn_enabled: true cudnn_benchmark: false cudnn_deterministic: false use_tf32: false 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: 4 nbest_averaging_interval: 0 grad_clip: 5.0 grad_clip_type: 2.0 grad_noise: false accum_grad: 4 no_forward_run: false resume: true train_dtype: float32 use_amp: true log_interval: 400 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: - frontend.upstream num_iters_per_epoch: null batch_size: 20 valid_batch_size: null batch_bins: 1200000 valid_batch_bins: null category_sample_size: 10 train_shape_file: - exp/asr_stats_raw_en_bpe1024/train/speech_shape - exp/asr_stats_raw_en_bpe1024/train/text_shape.bpe valid_shape_file: - exp/asr_stats_raw_en_bpe1024/valid/speech_shape - exp/asr_stats_raw_en_bpe1024/valid/text_shape.bpe batch_type: numel valid_batch_type: null fold_length: - 80000 - 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: - - dump/raw/train/wav.scp - speech - sound - - dump/raw/train/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: adam optim_conf: lr: 0.002 weight_decay: 1.0e-06 scheduler: warmuplr scheduler_conf: warmup_steps: 15000 token_list: - - - . - ▁I - ▁AND - '''' - ▁A - ▁YOU - S - ▁IT - T - ▁TO - ▁THE - ▁LIKE - ▁THAT - ▁NO - ▁BECAUSE - ▁ONE - ▁THEN - ▁DON - ▁TEETH - ▁TWO - ▁FIVE - ▁KNOW - ▁MY - ▁SO - ▁YOUR - ▁IS - ▁THEM - ▁DO - ▁SIX - ▁THREE - ▁G - ▁U - ▁TEN - ▁FOUR - ▁GET - ▁O - ▁K - ▁B - ▁L - ▁N - ▁S - ▁E - ▁M - ▁BRUSH - ▁THIS - ▁T - ▁CAN - ▁SEVEN - ▁EIGHT - ▁C - ▁HAVE - ▁PUT - ▁MAKE - ▁W - ▁J - ▁F - ▁IN - ▁P - ▁NINE - ▁Y - ▁D - ▁V - ▁OKAY - ▁Q - ▁Z - ▁ZERO - ▁IF - ▁H - ▁WHAT - ▁COUNT - ING - ▁R - ▁X - ▁OF - ▁HOW - ▁ - ▁WANT - ▁COLOR - ▁JUST - ▁WITH - ▁ON - N - ▁AN - ▁MIX - ▁COLORS - ▁THEY - ▁YEAH - ▁YES - ▁UP - ▁BLUE - ▁BY - ▁GO - M - ▁THERE - ▁ALL - ▁OR - ▁CLEAN - ED - ▁SEE - ▁BUT - ▁USE - ▁FOR - ▁BE - ▁TOOTHPASTE - ▁WAS - ▁UM - ▁LETTER - ▁NEED - ▁HE - ▁WILL - ▁PLUS - ▁DOG - ▁RED - RE - ▁PURPLE - ▁NOT - ▁CAVITIES - ▁OH - ▁ARE - ▁THINK - ▁WHY - ▁SHE - ▁DID - ▁HAT - Y - ▁PAINT - ▁BRUSHING - ▁BOX - ▁TOOTHBRUSH - ▁SICK - ▁OUT - ▁ME - ▁JUG - ▁DOES - ▁FLU - ▁MAKES - ▁WIG - ▁SH - ▁MAN - ▁WE - ▁MORE - OULD - ▁PLAY - ▁SOME - ▁JIBO - ▁GREEN - ▁VAN - ▁NUMBER - ▁YELLOW - ▁REALLY - D - ▁WHITE - ▁PINK - ▁WATER - ▁QUIZ - ▁NOW - ▁UH - ▁DIFFERENT - ▁RIGHT - IND - ▁SAY - ▁TREE - LL - CH - ▁HELP - ▁HUNDRED - ▁LOOK - ▁COULD - ▁COUNTING - ▁WAY - ▁MAYBE - ▁EASY - ▁WOULD - ▁BLACK - ▁TAKE - ▁HER - ▁LI - E - TTLE - F - ▁AL - ▁THING - ▁ELSE - ▁WELL - LY - ▁TOGETHER - ▁WHEN - ▁SIDE - ▁CAVITY - ▁FIRST - ▁DOWN - ▁DAY - ▁OTHER - ▁HERE - ▁CUBES - ▁COUNTED - ▁EVERY - ▁SA - ▁TELL - ▁DAD - ▁ORANGE - ▁SAME - ▁SOMETIMES - ▁MANY - OTHER - ID - ▁WON - ▁BIT - ▁HI - ▁TOO - ▁TIME - UH - ▁WAIT - ▁NOTHING - ▁FALL - ▁NAME - ▁LOT - ▁THAN - ▁EH - ▁MEAN - ▁NEW - W - H - ▁TOOTH - ER - ▁FLOSS - ▁START - ▁BROWN - ▁STACK - ▁NOPE - ▁GOOD - A - L - ▁LET - ▁WHI - O - ▁ALREADY - ▁INAUDIBLE - ▁MOUTH - ▁EAT - ▁HAS - ▁DONE - ▁THOSE - ▁BETTER - ▁FUN - ▁GERMS - TO - ▁UMM - CK - SO - EVEN - ▁WASH - ▁ACTUALLY - ▁DRINK - ▁FRIEND - ▁REMEMBER - ▁SUGAR - ▁SOMETHING - ▁HARD - ▁COME - ▁PAINTING - ▁SPI - ▁AT - I - TER - ▁MUCH - ▁GUESS - ▁HIM - ▁HA - IGHT - Z - ▁FRO - ▁IMPORTANT - ▁AGAIN - ▁STUFF - ▁BACK - ▁BUGS - ▁NIGHT - ▁ADD - G - ▁EA - HIS - K - EVER - ▁TH - ▁DARK - ▁FORGOT - ▁MOM - BODY - ▁UHHUH - ▁BAD - ▁TURN - ▁ANY - AH - EL - U - AKING - VERY - ▁GONNA - ▁FOUAH - ▁SURE - ▁PULL - ▁LONG - ▁KEEP - ES - P - ▁WAYS - TING - ALLY - VE - ONE - ▁QUESTION - ▁PAPER - ▁STU - YTHING - ▁SHOW - ▁CALLED - ▁LOVE - ▁MM - ▁TRY - ▁BYE - ▁TOP - LD - ▁MMM - ▁PE - ▁NUMBERS - BLE - PLE - ▁CUBE - OUT - R - ▁BOTTOM - ▁FAVORITE - ▁SPANISH - ▁TONGUE - ▁SCHOOL - ▁TWENTY - ▁MHM - ▁FRONT - ▁STAY - ▁SPELL - ▁TEEF - ▁LAST - ▁GUM - ▁HOLD - TY - ▁GROUPS - ▁OFF - ▁EQUALS - ▁FINGERS - ▁QUI - RAB - ▁MEANS - AW - ▁UHH - IT - WEE - ▁CH - ▁AM - ▁SI - RY - SIX - ▁WI - ▁BEAUTIFUL - ▁DENTIST - ▁HEALTHY - ▁HURT - ▁ZEWO - ▁KNEW - ▁MATH - ▁BOY - ▁HOLE - ▁DIRTY - ▁YET - ▁EX - ▁STARTED - ▁LIGHT - ▁THESE - ▁CU - B - ▁THINGS - ▁GRA - ▁WHO - ▁TWOS - ▁CIRCLE - ▁YO - ▁FINGER - ▁BA - CE - OTH - X - IR - MOST - ▁LEARN - FIVE - CI - ▁ANSWER - ▁EASIER - ▁LAUGHS - ▁MORNING - ▁MOUTHWASH - ▁PICTURE - ▁RINSE - ▁FORGET - ▁SISTER - ▁THOUGH - ▁TALKING - ▁GROW - ▁WHERE - ▁MINUTES - ▁SUP - ▁WISH - ▁OUR - ▁STI - ▁FLOSSING - SIC - EPT - ▁BIG - PER - ▁AH - TH - TEN - EN - ▁FAI - ▁ONES - ▁EQUAL - ▁SP - KAY - SIDE - WAYS - ▁AROUND - ▁PRETTY - ▁RAINBOW - ▁VIOLET - ▁LEFT - ▁GIRL - ▁SENSE - ▁SOUND - ▁EYES - ▁EVERYTHING - ▁GUY - ▁SHINY - ▁ELEVEN - ▁READY - ▁STICK - ▁FROG - ▁FOOD - ▁KEY - DE - ▁PL - ▁PART - OVE - ▁PR - ▁ROT - ▁TEE - ▁WERE - VER - ▁DIS - ▁HEY - USH - OH - IN - ISH - OVER - EEN - ▁MIND - ▁AB - SE - SH - DENTAL - OOL - ET - AR - ICK - NA - ENT - ▁BU - AT - UNTI - OW - OK - ▁EL - ▁MA - ▁QU - ▁WOR - ▁SIN - AKE - AND - ▁PRETEND - ▁BUS - ▁PLA - ▁CALL - ▁ONETWOTHREEFOUR - ▁CLASS - ▁CONNECT - ▁DISCOVER - ▁HOUSE - ▁RABBIT - ▁SQUEEZE - ▁THOUSAND - ▁ROBOT - ▁SCRUB - ▁SMELL - EXT - ▁BROTHER - ▁PILE - ▁BOTTLE - ▁PAINTBRUSH - IMA - ▁CROCODILE - ▁JUMP - ▁CANNOT - ▁TWICE - ▁STOP - UNCH - ▁SKIN - ▁TUR - ▁MOVING - IES - ▁FAST - ▁PRETENDING - EEP - ▁SHAKING - ▁MAY - ▁FAKE - ▁AWAY - ▁DI - ▁HAPP - ▁DUH - OO - ▁JUH - LE - ▁HUH - ▁BUH - BOOK - WENT - ▁CA - OSE - EM - IC - AG - ▁LETTERS - IS - EW - ONG - V - AL - PAY - REE - EE - ▁TIMES - ▁SPIN - UR - CU - GER - ▁TR - ▁AW - UGH - UT - ▁BL - ▁SL - ▁FORT - ▁GE - EA - ▁TA - GU - ▁FINISH - ▁UN - READ - THER - DAY - ▁BLA - ▁ARTIST - ▁BACKWARDS - ▁DOCTOR - ▁DREAMS - ▁EXPLA - ▁MIDDLE - ▁MOUSE - ▁PROB - ▁RINSING - ▁STRAIGHT - ▁SUNFLOWER - ▁TOOTHPICK - ▁TWELVE - ▁VULTURE - ▁CONFUS - TION - ▁HOME - ▁OPEN - ▁SORRY - ▁BORING - ▁MINE - ▁ENOUGH - ▁HELLO - ▁BORED - RITE - ▁TOWER - ▁BUIL - ▁ODD - ▁UNP - ▁APPLY - ▁ANYMORE - ▁FOUW - APE - OUNT - ▁FIFT - ▁ZEBRA - ▁LION - ▁BLAH - ▁BLOCK - ▁COP - ▁HMM - ▁ASK - ▁BAB - ▁DARKER - ▁HEAR - ▁CHO - ▁CLOSE - ▁JACK - ▁FULL - ▁CUP - ▁WHE - ▁IDEA - ▁PIRATES - ▁SPE - ▁HEAD - ▁GIVE - ▁END - DER - ▁HAND - ▁BOXES - ▁BEST - ▁LEARNING - ▁MESS - ▁MOST - ▁FLA - LIT - ▁AC - ▁AHW - ▁FUH - ▁LU - ▁SSS - OWN - ▁PUH - ▁PW - POS - ▁CIRCLES - ENS - LK - ▁PLAYING - AIL - AP - PIT - NG - ▁LETTU - IK - DDING - HH - PPER - ▁GW - ABL - OL - ▁KID - DING - ▁KA - ERS - ▁FI - LIP - ▁SE - ▁TREES - UN - ▁RO - ATE - ND - ▁FO - ICE - IF - HW - AY - ▁BIGGE - UST - ▁DE - ▁KI - ▁LOS - ▁THA - ▁PAN - IL - MB - ▁BOO - SPE - 'NO' - ACK - ▁FIN - C - ▁GROUP - ▁GERM - EAD - ▁SOMETIME - LZ - IVE - UP - TWO - HIRT - HRO - JELLYFISH - ▁PAR - PART - IBO - WHAT - KEY - FOUR - AME - ANGE - EC - TIME - ▁REAL - ELEPHANT - ▁BATHROOM - ▁BIVY - ▁BRACES - ▁FLOWER - ▁GARFIELD - ▁GARGLE - ▁KOALA - ▁PROBLEMS - ▁SEVENEIGHTNINE - ▁STINKY - ▁SWORD - ▁UPPERCASE - EMBER - FUL - ▁SEPARAT - ▁BEFORE - ▁BROKE - ▁LOUD - ▁MONSTER - ▁MOUF - ▁POOP - ▁SHINNY - ▁DRAW - ▁MAILBOX - ▁HUNGRY - ▁BREAK - ▁SARA - ▁JOB - ▁WATCH - ▁SPARKL - ▁SHORT - ▁WEEK - ▁BIRD - ▁MOMMY - ▁LOOSE - ▁GREAT - ▁PRETTIER - ▁SMIL - ▁FACE - ▁HAV - ▁PIECE - ▁FUNNY - ▁UNDER - ▁SLOWER - ACT - ▁PLEA - ▁VEHA - ▁PEAR - ▁FEEL - ▁SPIDER - ▁WORSE - ▁SWI - ▁AYE - UNU - ▁EVER - ▁HOPE - ▁SIGN - AK - UIZ - ▁SOFT - ▁POP - ▁TEEH - ▁DEH - IBLE - ▁SIDEWAYS - ROT - ▁ORDER - ▁FINISHED - ▁JELLYFISH - ▁FELL - KEU - ▁IMPO - HEAD - UM - ▁PRESS - ▁SECONDS - ▁LEA - ▁MOLD - LLUH - ▁READ - ▁ONETWO - ▁LINE - FE - ▁FOH - ▁HOT - ▁FOU - ▁MOH - ▁DEN - ▁WIN - ▁NINETY - IRTY - ▁TWEE - OUR - IRED - TLE - ▁HEH - ▁JU - PASTE - ▁FEVER - ▁WR - ▁PAI - MINT - TEEN - ▁WASHING - ▁BI - ▁NAH - DY - ▁RA - ▁DA - AHW - ▁YUH - ULL - ▁WL - UHTY - ▁SHO - ▁CUH - ASTE - OOD - ▁LAM - ▁CI - OLD - UNN - NUH - OCK - US - ▁SM - MPLE - ▁HIT - ▁THRO - ▁DEU - HOLE - ▁THINKING - UBB - ▁FU - ▁PI - ▁SMO - ▁VO - AN - UG - ▁HM - UE - GLE - ▁MOV - LI - ▁BLU - PORT - ▁WED - ▁TRI - ▁CHE - CA - ▁SC - ▁STO - ▁BED - ▁TELLS - ▁MI - OR - TTER - NES - OUGH - ▁AR - ROW - UA - AB - IG - OF - MAN - RK - OUN - ROUGH - LUH - DENT - ▁PIE - LAP - KUH - OT - RSE - ▁LA - ▁PAST - ▁ANOTH - OP - EP - ▁LATE - AM - LU - ▁WOO - HUH - ▁CER - OU - IPP - ▁CO - EH - TE - WHERE - ASH - PPY - WAY - RO - SHE - OST - AIN - ▁SECOND - ▁PIRATE - ▁MINUTE - ABET - ▁DIFFEREN - BE - IGH - ▁COO - ▁WHA - BRUSH - TRA - ▁PRES - ▁TRYIN - ▁GIV - OPE - SHIN - STRO - SIGN - ▁PLU - ZEBRA - LION - CROCODILE - LATE - UF - EQUALS - COME - UBE - J - TOGETHER - MAYBE - BOX - CLEAN - THEY - JIBO - EASY - MOUTH - ▁TALK - ▁SKI - WHY - WICE - CAUSE - UMP - TRIC - ▁CLOS - ▁SEVE - ▁DIRT - ▁NUMB - YOU - PRI - ▁JIB - ETTIER - FFERENT - ERCASE - ROOM - ▁DIFF - ▁JELLY - ▁SEVENEIGHT - ORGE - ▁YELL - DRA - ▁SLOW - ▁MON - ▁BUG - YPE - ▁BRU - COL - PUS - WO - INET - NGRY - BRUSHING - ▁CUB - OCTO - HIC - UDE - RUB - MOR - LOCK - ▁BR - YOUR - ▁STR - ▁KNE - ▁CRO - ▁BO - UALLY - ▁TOOTHB - ▁ANYMO - UKU - ▁GUE - MA - ENTY - PHA - ▁QUE - PF - KE - NOW - ▁LAS - ▁SHIN - ARN - GE - ▁MAIL - RUSHING - Q - init: null input_size: null 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: bpe bpemodel: data/en_token_list/bpe_unigram1024/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' short_noise_thres: 0.5 aux_ctc_tasks: [] frontend: s3prl frontend_conf: frontend_conf: upstream: wavlm_large download_dir: ./hub multilayer_feature: true 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 - 27 num_freq_mask: 2 apply_time_mask: true time_mask_width_ratio_range: - 0.0 - 0.05 num_time_mask: 5 normalize: utterance_mvn normalize_conf: {} model: espnet model_conf: ctc_weight: 0.3 lsm_weight: 0.1 length_normalized_loss: false extract_feats_in_collect_stats: false preencoder: linear preencoder_conf: input_size: 1024 output_size: 80 encoder: transformer encoder_conf: output_size: 256 attention_heads: 4 linear_units: 1024 num_blocks: 18 dropout_rate: 0.1 positional_dropout_rate: 0.1 attention_dropout_rate: 0.1 input_layer: conv2d2 normalize_before: true postencoder: null postencoder_conf: {} 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: 0.1 src_attention_dropout_rate: 0.1 preprocessor: default preprocessor_conf: {} required: - output_dir - token_list version: '202402' 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} } ```