We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Sparse Distillation: Speeding Up Text Classification by Using Bigger Models

Abstract: Distilling state-of-the-art transformer models into lightweight student models is an effective way to reduce computation cost at inference time. However, the improved inference speed may be still unsatisfactory for certain time-sensitive applications. In this paper, we aim to further push the limit of inference speed by exploring a new area in the design space of the student model. More specifically, we consider distilling a transformer-based text classifier into a billion-parameter, sparsely-activated student model with a embedding-averaging architecture. Our experiments show that the student models retain 97% of the RoBERTa-Large teacher performance on a collection of six text classification tasks. Meanwhile, the student model achieves up to 600x speed-up on both GPUs and CPUs, compared to the teacher models. Further investigation shows that our pipeline is also effective in privacy-preserving and domain generalization settings.
Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG)
Cite as: arXiv:2110.08536 [cs.CL]
  (or arXiv:2110.08536v1 [cs.CL] for this version)

Submission history

From: Qinyuan Ye [view email]
[v1] Sat, 16 Oct 2021 10:04:14 GMT (166kb,D)

Link back to: arXiv, form interface, contact.