WavLM: Large-Scale Self-Supervised Pre-training for Full Stack Speech Processing

Official PyTorch implementation and pretrained models of WavLM

• Dec 2021: An interesting speaker verification demo on HuggingFace. You can have a try!
• Dec 2021: WavLM Large Release and HuggingFace Support
• Nov 2021: release code and pretrained models (WavLM Base and WavLM Base+)
• Oct 2021: release preprint in arXiv

import torch
from WavLM import WavLM, WavLMConfig

# load the pre-trained checkpoints
checkpoint = torch.load('/path/to/wavlm.pt')
cfg = WavLMConfig(checkpoint['cfg'])
model = WavLM(cfg)
model.load_state_dict(checkpoint['model'])
model.eval()

# extract the representation of last layer
wav_input_16khz = torch.randn(1,10000)
if cfg.normalize:
    wav_input_16khz = torch.nn.functional.layer_norm(wav_input_16khz , wav_input_16khz.shape)
rep = model.extract_features(wav_input_16khz)[0]

# extract the representation of each layer
wav_input_16khz = torch.randn(1,10000)
if cfg.normalize:
    wav_input_16khz = torch.nn.functional.layer_norm(wav_input_16khz , wav_input_16khz.shape)
rep, layer_results = model.extract_features(wav_input_16khz, output_layer=model.cfg.encoder_layers, ret_layer_results=True)[0]
layer_reps = [x.transpose(0, 1) for x, _ in layer_results]

HuggingFace and s3prl both support our models. It is very easy to fine-tune our models on different downstream tasks. We suggest you to extract representation of each layer and weighted sum the representations.

We also evaluate our models on typical speech processing benchmarks.

Finetune the model with VoxCeleb2 dev data, and evaluate it on the VoxCeleb1

Evaluation on the LibriCSS

Evaluation on the CALLHOME

Evaluate on the LibriSpeech

Please visit here for more interesting and effective pre-trained models

This project is licensed under the license found in the LICENSE file in the root directory of this source tree. Portions of the source code are based on the FAIRSEQ project.

Microsoft Open Source Code of Conduct

If you find our work is useful in your research, please cite the following paper:

@article{Chen2021WavLM,
  title   = {WavLM: Large-Scale Self-Supervised  Pre-training   for Full Stack Speech Processing},
  author  = {Sanyuan Chen and Chengyi Wang and Zhengyang Chen and Yu Wu and Shujie Liu and Zhuo Chen and Jinyu Li and Naoyuki Kanda and Takuya Yoshioka and Xiong Xiao and Jian Wu and Long Zhou and Shuo Ren and Yanmin Qian and Yao Qian and Jian Wu and Michael Zeng and Furu Wei},
  eprint={2110.13900},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  year={2021}
}

For help or issues using WavLM models, please submit a GitHub issue.

For other communications related to WavLM, please contact Yu Wu ([email protected]).