AI Paper Atlas

Research map / Architectures and efficiency / Quantization

ActNN: Reducing Training Memory Footprint via 2-Bit Activation Compressed Training

Jianfei Chen, Lianmin Zheng, Zhewei Yao, Dequan Wang, Ion Stoica, Michael Mahoney, Joseph Gonzalez

ICML 2021 · Architectures and efficiency › Quantization · approach: Low-precision training formats · 10 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
Use ActNN (L3) activation compression at 2 bits instead of full-precision activationsResNet-50 · ImageNettop-1 accuracy: 77.1 -> 76.9No clear effect
Replace BLPA per-tensor quantization with ActNN (L3) mixed-precision quantization at 1.5 bitsResNet-50 · ImageNettop-1 accuracy: 76.6 -> 76.4Mixed
Allocate bits across layers as well as samples (L3 rather than L2.5) (ablation)ActNN quantization · ImageNet, ResNet-50top-1 accuracy: 75.9 -> 76.4Improved
Allocate bits between samples while keeping layer precision equal (L2.5 instead of L2) (ablation)ActNN quantization · ImageNet, ResNet-50 at 2 bitstop-1 accuracy: 0.1 -> 77.1No clear effect
Combine ActNN (L3) activation compression with AMPResNet-50 · ImageNetvalidation accuracy: 77.3 -> 76.9No clear effect
Increase the batch size from 8 to 160 with ActNN (L3)Dilation8 · Semantic segmentationmIoU: 72.92 -> 75.85Improved
Increase the batch size from 8 to 160 with ActNN (L3)FPN · Semantic segmentationmIoU: 74.18 -> 75.98Improved
Use ActNN (L3), 2-bit activations, and a larger batchMoCov2 · ImageNetvalidation accuracy: 67.69 -> 67.25No clear effect
Use ActNN (L3), 2-bit activations, and a larger batchBYOL · ImageNetvalidation accuracy: 72.35 -> 72.65Improved
Estimate gradient magnitude using moving averages rather than stale gradients (ablation)ActNN (L3) · Quantization strategy ablationaccuracyNo clear effect

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