File size: 3,112 Bytes
8ddc011 e763a34 8ddc011 5d6eb67 8ddc011 5d6eb67 8ddc011 5d6eb67 8ddc011 5d6eb67 8ddc011 5d6eb67 8ddc011 5d6eb67 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 | ---
license: cc-by-nc-sa-4.0
pipeline_tag: feature-extraction
tags:
- automatic-speech-recognition
- audio-classification
- audio
- speech
- music
library_name: transformers
datasets:
- openslr/librispeech_asr
- facebook/multilingual_librispeech
- mozilla-foundation/common_voice_17_0
- speechcolab/gigaspeech
- facebook/voxpopuli
- agkphysics/AudioSet
language:
- en
---
# USAD: Universal Speech and Audio Representation via Distillation
**Universal Speech and Audio Distillation (USAD)** is a unified **speech**, **sound**, and **music** encoder distilled from domain-specific teachers.
Trained on 126k hours of mixed data, USAD delivers competitive performance across diverse benchmarks (SUPERB, HEAR, and AudioSet) with a single model.
[👀 **Read Full Paper**](https://arxiv.org/abs/2506.18843)
[🛠️ **GitHub**](https://github.com/vectominist/usad)
---
## 🗂️ Models
USAD models are all transformer encoders operating at **50Hz frame rate**. The teacher models are **WavLM Base+** and **ATST Frame**.
| Model | Parameters | Dim | Layer |
| :-------------------------------------------------------- | ---------: | ---: | ----: |
| [USAD Small](https://huggingface.co/MIT-SLS/USAD-Small) | 24M | 384 | 12 |
| [USAD Base](https://huggingface.co/MIT-SLS/USAD-Base) | 94M | 768 | 12 |
| [USAD Large]((https://huggingface.co/MIT-SLS/USAD-Small)) | 330M | 1024 | 24 |
---
## 🚀 How To Use
**Installation**
```
pip install -U torch torchaudio transformers
```
**Load Model and Extract Features**
```python
import torch
from transformers import AutoModel
# Load pre-trained model
model = AutoModel.from_pretrained("MIT-SLS/USAD-Base", trust_remote_code=True).cuda().eval()
# Load audio and resample to 16kHz
wav = model.load_audio("path/to/audio").unsqueeze(0) # (batch_size, wav_len)
# wav is a float tensor on the same device as the model
# You can also load waveforms directly with torchaudio.load
# Extract features
with torch.no_grad():
results = model(wav)
# result["x"]: model final output (batch_size, seq_len)
# result["mel"]: mel fbank (batch_size, seq_len * 2, mel_dim)
# result["hidden_states"]: list of (batch_size, seq_len, encoder_dim)
# result["ffn"]: list of (batch_size, seq_len, encoder_dim)
```
See [usad_model.py](https://huggingface.co/MIT-SLS/USAD-Base/blob/main/usad_model.py) for more details about the model.
---
## 📖 Citation
```bibtex
@inproceedings{chang2025usad,
title={{USAD}: Universal Speech and Audio Representation via Distillation},
author={Chang, Heng-Jui and Bhati, Saurabhchand and Glass, James and Liu, Alexander H.},
booktitle={IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)},
year={2025}
}
```
---
## 🙏 Acknowledgement
Our implementation is based on the awesome [facebookresearch/fairseq](https://github.com/facebookresearch/fairseq), [cwx-worst-one/EAT](https://github.com/cwx-worst-one/EAT), and [sooftware/conformer](https://github.com/sooftware/conformer) repositories. |