AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Deep Compression of Pre-trained Transformer Models

Naigang Wang, Chi-Chun Liu, Swagath Venkataramani, Sanchari Sen, Chia-Yu Chen, Kaoutar El Maghraoui, Viji Srinivasan, Leland Chang

NeurIPS 2022 · Architectures and efficiency › Pruning and efficient tuning · approach: Efficient ViT adaptation · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add zero alignment to weight quantization (ablation)sparse INT4 models · BERT-base SQuAD1.1F1 (%): 86.06 -> 87.07Improved
add zero alignment to weight quantization (ablation)sparse INT4 models · Wav2vec2.0 LibrispeechWER (%): 5.76 -> 4.53Improved
add zero alignment to weight quantization (ablation)sparse INT4 models · ViT ImageNet1kaccuracy (%): 83.25 -> 83.60Improved
use PACT instead of MinMax activation quantizer (ablation)INT8 models · BERT-base SQuAD1.1F1 (%): 88.35 -> 87.89Hurt
use PACT instead of MinMax activation quantizer (ablation)INT8 models · Wav2vec2.0 LibrispeechWER (%): 3.85 -> 4.05Hurt
use PACT instead of MinMax activation quantizer (ablation)INT8 models · ViT ImageNet1kaccuracy (%): 84.47 -> 83.98Hurt
initialize INT4 from fine-tuned FP32 instead of pre-trained model (ablation)INT4 models · BERT-base SQuAD1.1F1 (%): 86.68 -> 87.86Improved
initialize INT4 from fine-tuned FP32 instead of pre-trained model (ablation)INT4 models · Wav2vec2.0 LibrispeechWER (%): 5.02 -> 4.56Improved
initialize INT8 from fine-tuned FP32 instead of pre-trained model (ablation)INT8 models · ViT ImageNet1kaccuracy (%): 83.05 -> 84.47Improved
initialize sparse INT4 from sparse INT8 instead of FP32 fine-tuned model (ablation)sparse INT4 models · BERT-base SQuAD1.1F1 (%): 86.75 -> 87.07Improved
initialize sparse INT4 from sparse INT8 instead of FP32 fine-tuned model (ablation)sparse INT4 models · Wav2vec2.0 LibrispeechWER (%): 5.09 -> 4.65Improved
initialize sparse INT4 from sparse INT8 instead of FP32 fine-tuned model (ablation)sparse INT4 models · ViT ImageNet1kaccuracy (%): 82.66 -> 83.60Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy