--- tags: - espnet - audio - spoken-language-understanding language: en datasets: - slue-ted license: cc-by-4.0 --- ## ESPnet2 SLU model ### `espnet/slueted_whisper_summ` This model was trained by “siddhu001” using slue-ted 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 e23ef85f0b3116ad5c60d0833f186da0deec0734 pip install -e . cd egs2/slue-ted/slu1 ./run.sh --skip_data_prep false --skip_train true --download_model espnet/slueted_whisper_summ ``` {'rouge1': 0.2255418629519756, 'rouge2': 0.0485061537185737, 'rougeL': 0.1596465851004139, 'rougeLsum': 0.15968116069467322, 'meteor': 0.2129616261465529} RESULT 22.55418629519756 3.799127541421444e-132 15.96465851004139 21.29616261465529 83.78519008627457 ## SLU config
expand ``` config: conf//train_asr_whisper_weighted_conv2d2.yaml print_config: false log_level: INFO drop_last_iter: false dry_run: false iterator_type: sequence valid_iterator_type: null output_dir: exp/slu_train_asr_whisper_weighted_conv2d2_raw_en_bpe500_sp ngpu: 1 seed: 2022 num_workers: 2 num_att_plot: 3 dist_backend: nccl dist_init_method: env:// dist_world_size: 4 dist_rank: 0 local_rank: 0 dist_master_addr: localhost dist_master_port: 42635 dist_launcher: null multiprocessing_distributed: true unused_parameters: false sharded_ddp: false cudnn_enabled: true cudnn_benchmark: false cudnn_deterministic: true collect_stats: false write_collected_feats: false max_epoch: 25 patience: null val_scheduler_criterion: - valid - loss early_stopping_criterion: - valid - loss - min best_model_criterion: - - valid - acc - max keep_nbest_models: 10 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: 100 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_lora: false save_lora_only: true lora_conf: {} pretrain_path: null init_param: - /scratch/bbjs/arora1/espnet_slue_PR/espnet/egs2/tedlium3/asr1/exp/asr_train_asr_whisper_weighted_conv2d2_raw_en_bpe500/valid.acc.ave_10best.pth:::ctc ignore_init_mismatch: false freeze_param: - encoder num_iters_per_epoch: null batch_size: 20 valid_batch_size: null batch_bins: 12000000 valid_batch_bins: null train_shape_file: - exp/slu_stats_raw_en_bpe500_sp/train/speech_shape - exp/slu_stats_raw_en_bpe500_sp/train/text_shape.bpe valid_shape_file: - exp/slu_stats_raw_en_bpe500_sp/valid/speech_shape - exp/slu_stats_raw_en_bpe500_sp/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 train_data_path_and_name_and_type: - - dump/raw/train_sp/wav.scp - speech - kaldi_ark - - dump/raw/train_sp/text - text - text valid_data_path_and_name_and_type: - - dump/raw/devel/wav.scp - speech - kaldi_ark - - dump/raw/devel/text - text - text 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: 5000 token_list: - - - '[sep]' - '"' - s - ▁ - ▁the - ',' - t - d - ▁a - . - ing - o - e - ▁to - a - ▁and - y - n - ▁of - r - ▁in - u - i - m - p - c - er - g - l - al - re - ed - b - '''' - ar - k - in - f - ▁" - le - 'on' - v - or - th - '-' - ▁c - en - ▁f - ▁-- - ▁we - ▁for - ▁how - ly - ▁re - se - ▁that - es - w - ic - st - ▁w - ▁be - ri - an - ra - ve - ce - ur - ▁by - ▁it - li - ▁de - '?' - it - ch - ent - ▁is - ter - el - ▁on - ▁e - ▁he - ▁co - ▁an - ▁ma - ▁st - ll - ▁with - ▁can - il - ▁you - ▁us - ation - te - ▁this - ▁b - ▁do - ▁g - me - ▁what - ck - ▁from - ate - ▁p - z - la - ▁mo - ▁di - ive - mp - ▁talk - ity - vi - ta - at - ge - ▁tr - ▁she - ▁our - ▁pa - ci - et - h - ▁su - ver - ▁world - pe - ▁about - ▁me - ▁so - and - ▁con - tion - de - ir - ▁her - im - ':' - ▁his - ies - ▁po - ▁are - ect - lo - ▁your - un - ist - hi - ▁mi - x - id - ment - ol - ul - ti - ne - qu - ▁but - ▁ca - ▁fa - ▁as - ▁un - ers - ight - ▁says - '0' - ng - op - '1' - ▁k - ad - j - ma - ▁pro - ▁work - ▁ba - ▁share - ▁new - ▁more - ▁vi - ▁sa - ▁at - ▁la - ut - bi - sion - ▁ho - na - act - age - ke - if - ▁bo - ▁br - ▁ha - ▁no - co - ▁lo - mi - ▁make - ▁people - ▁why - ant - ▁their - ▁i - ▁life - ▁all - ting - ▁human - ▁have - om - ) - ▁( - ▁help - ▁ted - wa - sh - ▁da - ▁le - ▁out - ph - ical - ▁way - ff - ▁ro - able - ▁some - est - ure - em - ho - ▁ex - gen - ha - ia - ine - ▁into - ca - ▁was - ▁who - ther - ▁they - ow - he - ▁one - ▁when - form - ▁pre - ni - ▁could - ▁like - ▁per - ▁up - ance - com - ▁go - ion - tor - ▁fe - ▁ra - ▁or - ▁en - ▁change - tic - ▁every - ▁jo - ence - ▁not - ▁art - one - use - ous - ▁plan - ▁music - ▁exp - und - ▁ne - um - ative - pp - ▁need - tro - directed - ▁learn - ▁narrate - ▁has - lar - '].' - man - ▁car - ▁future - ▁real - ▁time - ize - ▁live - ber - ▁mar - ▁ga - ▁take - ▁dr - ful - ▁get - ▁shows - day - ▁cha - ▁than - ▁know - ian - ▁see - ▁just - '2' - ▁other - old - ▁design - ▁chi - ▁build - ious - ▁most - ▁si - ▁will - ▁power - ▁think - port - ▁over - ▁ja - ish - ▁climate - ▁sha - ▁through - less - '3' - ▁my - ▁where - ▁global - ▁health - ▁pri - ▁20 - ▁story - gu - ugh - ▁create - ▁look - ▁trans - ▁har - ▁even - ▁part - ▁years - ▁lead - side - low - long - ▁technolog - ness - '5' - ▁call - ▁sc - ▁system - '9' - line - ▁brain - ▁data - ▁own - ition - ▁explains - ▁tell - ▁explore - ▁start - ▁ru - ▁which - ▁anderson - ▁find - ▁hu - ▁women - ▁better - ▁idea - ▁history - ▁research - ▁science - ism - ▁first - ▁grow - ▁right - clu - ▁space - ▁develop - ▁problem - ▁two - ▁earth - ologist - ▁many - ▁should - ▁three - ▁fellow - ▁social - ▁africa - ▁... - '4' - ▁addis - ▁powerful - ▁found - ▁under - ▁understand - ▁after - ▁stories - ▁around - ▁personal - ▁project - ▁between - ▁question - ▁play - ▁scientist - ▁happen - ▁good - ▁produc - ▁experience - ▁step - ▁america - '8' - ▁great - ▁down - ▁high - ▁would - ▁turn - ▁surprising - ▁imagin - ▁teach - cross - ▁place - ▁medic - ▁million - ▁things - '7' - ▁reveal - ▁without - ▁challenge - ▁next - ▁each - ▁studio - organ - '6' - ▁business - ▁much - ▁show - ▁conversation - ▁energy - ▁school - ▁ocean - ▁while - source - ization - ▁break - ▁robot - ▁disease - ▁behind - ability - ▁team - ▁chris - ▁become - ▁solution - ▁protect - ▁collect - ▁different - ▁those - ▁connect - ▁architect - ▁language - ▁simple - ▁solve - ▁before - ▁community - ▁country - ▁secret - ▁keep - ▁food - ▁thought - ▁discover - ▁environment - ▁government - ▁public - ; - '!' - / - q - '%' - '@' - ']' - + - '&' - '|' - _ - ( - '"' - $ - '*' - '=' - '[' - '`' - transcript_token_list: null two_pass: false pre_postencoder_norm: false init: null 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 token_type: bpe bpemodel: data/en_token_list/bpe_unigram500/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 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: false weighted_sum: true extract_feats_in_collect_stats: false preencoder: null preencoder_conf: {} encoder: whisper encoder_conf: whisper_model: medium dropout_rate: 0.0 use_specaug: true specaug_conf: apply_time_warp: true time_warp_window: 5 time_warp_mode: bicubic apply_freq_mask: true freq_mask_width_range: - 0 - 40 num_freq_mask: 2 apply_time_mask: true time_mask_width_ratio_range: - 0.0 - 0.12 num_time_mask: 5 prepostencoder: linear prepostencoder_conf: input_size: 1024 output_size: 80 postencoder: conformer_full postencoder_conf: output_size: 256 attention_heads: 4 linear_units: 1024 num_blocks: 12 dropout_rate: 0.1 positional_dropout_rate: 0.1 attention_dropout_rate: 0.1 input_layer: conv2d2 normalize_before: true macaron_style: true rel_pos_type: latest pos_enc_layer_type: rel_pos selfattention_layer_type: rel_selfattn activation_type: swish use_cnn_module: true cnn_module_kernel: 31 deliberationencoder: null deliberationencoder_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 postdecoder: null postdecoder_conf: {} required: - output_dir - token_list version: '202310' distributed: true ```
### Citing ESPnet ```BibTex @inproceedings{ESPnet-SLU, title={{ESPnet-SLU}: Advancing spoken language understanding through espnet}, author={Arora, Siddhant and Dalmia, Siddharth and Denisov, Pavel and Chang, Xuankai and Ueda, Yushi and Peng, Yifan and Zhang, Yuekai and Kumar, Sujay and Ganesan, Karthik and Yan, Brian and others}, booktitle={ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, pages={7167--7171}, year={2022}, organization={IEEE} } @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{arora2021espnet, title={ESPnet-SLU: Advancing Spoken Language Understanding through ESPnet}, author={Arora, Siddhant and Dalmia, Siddharth and Denisov, Pavel and Chang, Xuankai and Ueda, Yushi and Peng, Yifan and Zhang, Yuekai and Kumar, Sujay and Ganesan, Karthik and Yan, Brian and others}, eprint={2111.14706}, archivePrefix={arXiv}, primaryClass={cs.CL} year={2021} } @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} } ```