Pretrained Transformers as Universal Computation Engines

03/09/2021
by   Kevin Lu, et al.
0

We investigate the capability of a transformer pretrained on natural language to generalize to other modalities with minimal finetuning – in particular, without finetuning of the self-attention and feedforward layers of the residual blocks. We consider such a model, which we call a Frozen Pretrained Transformer (FPT), and study finetuning it on a variety of sequence classification tasks spanning numerical computation, vision, and protein fold prediction. In contrast to prior works which investigate finetuning on the same modality as the pretraining dataset, we show that pretraining on natural language improves performance and compute efficiency on non-language downstream tasks. In particular, we find that such pretraining enables FPT to generalize in zero-shot to these modalities, matching the performance of a transformer fully trained on these tasks.

READ FULL TEXT

page 1

page 7

06/29/2020

Knowledge-Aware Language Model Pretraining

How much knowledge do pretrained language models hold? Recent research o...
08/08/2022

Investigating Efficiently Extending Transformers for Long Input Summarization

While large pretrained Transformer models have proven highly capable at ...
10/13/2021

Study of positional encoding approaches for Audio Spectrogram Transformers

Transformers have revolutionized the world of deep learning, specially i...
10/18/2021

NormFormer: Improved Transformer Pretraining with Extra Normalization

During pretraining, the Pre-LayerNorm transformer suffers from a gradien...
10/09/2021

Vector-quantized Image Modeling with Improved VQGAN

Pretraining language models with next-token prediction on massive text c...
12/31/2020

Verb Knowledge Injection for Multilingual Event Processing

In parallel to their overwhelming success across NLP tasks, language abi...
10/03/2020

Mining Knowledge for Natural Language Inference from Wikipedia Categories

Accurate lexical entailment (LE) and natural language inference (NLI) of...

Code Repositories

universal-computation

Official codebase for Pretrained Transformers as Universal Computation Engines.


view repo