BEATs-base-AS2M-iter1

BEATs base encoder — AudioSet-2M, iteration 1

BEATs base audio encoder pre-trained on AudioSet-2M with ESPnet, reproducing BEATs (Table 4). Iteration 1 is trained against the discrete targets of the iteration-1 VQ tokenizer. Produced by the egs2/audioset/ssl1 recipe.

Usage

Load into an ESPnet downstream classification recipe (e.g. egs2/esc50/asr1, egs2/as20k/cls1) via beats_ckpt_path:

encoder: beats
encoder_conf:
    beats_ckpt_path: beats_encoder_iter1.pt
    fbank_mean: 15.66439
    fbank_std: 6.38312

Downstream results

Citation

@inproceedings{chen2022beats,
  title={BEATs: Audio Pre-Training with Acoustic Tokenizers},
  author={Chen, Sanyuan and Wu, Yu and Wang, Chengyi and Liu, Shujie and
          Tompkins, Daniel and Chen, Zhuo and Wei, Furu},
  booktitle={ICML}, year={2023}
}
@inproceedings{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},
  booktitle={WASPAA}, year={2025}
}