Task-Agnostic Structured Pruning of Speech Representation Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2023
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910930373181440 |
|---|---|
| author | Wang, Haoyu Wang, Siyuan Zhang, Wei-Qiang Suo, Hongbin Wan, Yulong |
| author_facet | Wang, Haoyu Wang, Siyuan Zhang, Wei-Qiang Suo, Hongbin Wan, Yulong |
| contents | Self-supervised pre-trained models such as Wav2vec2, Hubert, and WavLM have been shown to significantly improve many speech tasks. However, their large memory and strong computational requirements hinder their industrial applicability. Structured pruning is a hardware-friendly model compression technique but usually results in a larger loss of accuracy. In this paper, we propose a fine-grained attention head pruning method to compensate for the performance degradation. In addition, we also introduce the straight through estimator into the L0 regularization to further accelerate the pruned model. Experiments on the SUPERB benchmark show that our model can achieve comparable performance to the dense model in multiple tasks and outperforms the Wav2vec 2.0 base model on average, with 72% fewer parameters and 2 times faster inference speed. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2306_01385 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Task-Agnostic Structured Pruning of Speech Representation Models Wang, Haoyu Wang, Siyuan Zhang, Wei-Qiang Suo, Hongbin Wan, Yulong Audio and Speech Processing Computation and Language Sound Self-supervised pre-trained models such as Wav2vec2, Hubert, and WavLM have been shown to significantly improve many speech tasks. However, their large memory and strong computational requirements hinder their industrial applicability. Structured pruning is a hardware-friendly model compression technique but usually results in a larger loss of accuracy. In this paper, we propose a fine-grained attention head pruning method to compensate for the performance degradation. In addition, we also introduce the straight through estimator into the L0 regularization to further accelerate the pruned model. Experiments on the SUPERB benchmark show that our model can achieve comparable performance to the dense model in multiple tasks and outperforms the Wav2vec 2.0 base model on average, with 72% fewer parameters and 2 times faster inference speed. |
| title | Task-Agnostic Structured Pruning of Speech Representation Models |
| topic | Audio and Speech Processing Computation and Language Sound |
| url | https://arxiv.org/abs/2306.01385 |