README.md

---
tags:
- espnet
- audio
- classification
datasets:
- nsynth
license: cc-by-4.0
---

## ESPnet2 CLS model

### `espnet/OpenBEATS-Large-NsynthPitch`

This model was trained by Shikhar Bharadwaj using nsynth recipe in [espnet](https://github.com/espnet/espnet/).

## CLS config

<details><summary>expand</summary>

```
config: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/earlarge2/conf/ear_large/nsynth_pitch.yaml
print_config: false
log_level: INFO
drop_last_iter: false
dry_run: false
iterator_type: sequence
valid_iterator_type: null
output_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/nsynth_pitch/cls_earlarge2
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
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: 30
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: audioverse
wandb_id: null
wandb_entity: shikhar
wandb_name: nsynth_pitch.earlarge2
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: null
batch_size: 32
valid_batch_size: 16
batch_bins: 1000000
valid_batch_bins: null
category_sample_size: 10
train_shape_file:
- /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/nsynth_pitch/cls_stats_16k/train/speech_shape
- /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/nsynth_pitch/cls_stats_16k/train/label_shape
valid_shape_file:
- /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/nsynth_pitch/cls_stats_16k/valid/speech_shape
- /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/nsynth_pitch/cls_stats_16k/valid/label_shape
batch_type: folded
valid_batch_type: null
fold_length:
- 480000
- 600
sort_in_batch: descending
shuffle_within_batch: false
sort_batch: descending
multiple_iterator: false
utt2weight_file: null
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:
-   - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/nsynth_pitch/train/wav.scp
    - speech
    - sound
-   - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/nsynth_pitch/train/text
    - label
    - text
valid_data_path_and_name_and_type:
-   - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/nsynth_pitch/valid/wav.scp
    - speech
    - sound
-   - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/nsynth_pitch/valid/text
    - label
    - 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: 3.0e-05
    weight_decay: 0.01
    betas:
    - 0.9
    - 0.98
scheduler: cosineannealingwarmuprestarts
scheduler_conf:
    first_cycle_steps: 50000
    warmup_steps: 5000
    max_lr: 3.0e-05
    min_lr: 5.0e-06
lightning_conf: {}
token_list:
- '60'
- '55'
- '62'
- '63'
- '61'
- '57'
- '56'
- '59'
- '58'
- '64'
- '53'
- '65'
- '67'
- '52'
- '54'
- '69'
- '70'
- '72'
- '66'
- '68'
- '71'
- '50'
- '74'
- '51'
- '48'
- '73'
- '76'
- '49'
- '75'
- '77'
- '79'
- '78'
- '81'
- '46'
- '80'
- '82'
- '47'
- '44'
- '45'
- '84'
- '42'
- '83'
- '43'
- '41'
- '40'
- '39'
- '38'
- '37'
- '36'
- '85'
- '86'
- '35'
- '34'
- '33'
- '32'
- '31'
- '88'
- '87'
- '30'
- '89'
- '29'
- '28'
- '27'
- '25'
- '26'
- '24'
- '91'
- '90'
- '92'
- '93'
- '94'
- '95'
- '96'
- '23'
- '22'
- '97'
- '98'
- '99'
- '100'
- '21'
- '101'
- '103'
- '102'
- '104'
- '105'
- '106'
- '108'
- '107'
- '17'
- '19'
- '16'
- '12'
- '10'
- '18'
- '14'
- '20'
- '11'
- '13'
- '9'
- '15'
- '109'
- '110'
- '115'
- '114'
- '117'
- '111'
- '112'
- '118'
- '116'
- '113'
- '120'
- '119'
- <blank>
- <unk>
text_token_list: null
text_bpemodel: null
init: xavier_normal
input_size: 1
use_preprocessor: true
frontend: null
frontend_conf: {}
specaug: null
specaug_conf: {}
normalize: null
normalize_conf: {}
preencoder: null
preencoder_conf: {}
encoder: beats
encoder_conf:
    beats_ckpt_path: /work/nvme/bbjs/sbharadwaj/7Msounds/exp/beats_iter1_large1.tune_lr1.0e-4_warmup40000_bins1600000_totalsteps400000/epoch_latest.pt
    beats_config:
        layer_wise_gradient_decay_ratio: 0.3
        encoder_layerdrop: 0.1
        dropout: 0.0
    use_weighted_representation: false
    specaug_config:
        apply_time_warp: true
        apply_freq_mask: false
        apply_time_mask: true
        time_mask_width_ratio_range:
        - 0
        - 0.06
        num_time_mask: 1
    roll_augment: true
    roll_interval: 1
text_encoder: null
text_encoder_conf: {}
embedding_fusion: null
embedding_fusion_conf: {}
decoder: linear
decoder_conf: {}
model: espnet
model_conf:
    classification_type: multi-class
    lsm_weight: 0.1
required:
- output_dir
- token_list
version: '202412'
distributed: false
```

</details>

### Citations

```BibTex

@article{bharadwaj2025openbeats,
  title={OpenBEATs: A Fully Open-Source General-Purpose Audio Encoder},
  author={Bharadwaj, Shikhar and Cornell, Samuele and Choi, Kwanghee and Fukayama, Satoru and Shim, Hye-jin and Deshmukh, Soham and Watanabe, Shinji},
  journal={arXiv preprint arXiv:2507.14129},
  year={2025}
}

@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}
}






```