GenDistiller: Distilling Pre-trained Language Models based on an Autoregressive Generative Model
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909228618219520 |
|---|---|
| author | Gao, Yingying Zhang, Shilei Deng, Chao Feng, Junlan |
| author_facet | Gao, Yingying Zhang, Shilei Deng, Chao Feng, Junlan |
| contents | Pre-trained speech language models such as HuBERT and WavLM leverage unlabeled speech data for self-supervised learning and offer powerful representations for numerous downstream tasks. Despite the success of these models, their high requirements for memory and computing resource hinder their application on resource restricted devices. Therefore, this paper introduces GenDistiller, a novel knowledge distillation framework which generates the hidden representations of the pre-trained teacher model directly by a much smaller student network. The proposed method takes the previous hidden layer as history and implements a layer-by-layer prediction of the teacher model autoregressively. Experiments on SUPERB reveal the advantage of GenDistiller over the baseline distilling method without an autoregressive framework, with 33% fewer parameters, similar time consumption and better performance on most of the SUPERB tasks. Ultimately, the proposed GenDistiller reduces the size of WavLM by 82%. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_09444 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | GenDistiller: Distilling Pre-trained Language Models based on an Autoregressive Generative Model Gao, Yingying Zhang, Shilei Deng, Chao Feng, Junlan Audio and Speech Processing Computation and Language Sound Pre-trained speech language models such as HuBERT and WavLM leverage unlabeled speech data for self-supervised learning and offer powerful representations for numerous downstream tasks. Despite the success of these models, their high requirements for memory and computing resource hinder their application on resource restricted devices. Therefore, this paper introduces GenDistiller, a novel knowledge distillation framework which generates the hidden representations of the pre-trained teacher model directly by a much smaller student network. The proposed method takes the previous hidden layer as history and implements a layer-by-layer prediction of the teacher model autoregressively. Experiments on SUPERB reveal the advantage of GenDistiller over the baseline distilling method without an autoregressive framework, with 33% fewer parameters, similar time consumption and better performance on most of the SUPERB tasks. Ultimately, the proposed GenDistiller reduces the size of WavLM by 82%. |
| title | GenDistiller: Distilling Pre-trained Language Models based on an Autoregressive Generative Model |
| topic | Audio and Speech Processing Computation and Language Sound |
| url | https://arxiv.org/abs/2406.09444 |