AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

TinyTL: Reduce Memory, Not Parameters for Efficient On-Device Learning

Han Cai, Chuang Gan, Ligeng Zhu, Song Han

NeurIPS 2020 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 44 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 lite residual modules to bias-only tuning (ablation)TinyTL-B · Carstop-1 accuracy (%): 73.4 -> 85.0Improved
add lite residual modules to bias-only tuning (ablation)TinyTL-B · Aircrafttop-1 accuracy (%): 63.2 -> 75.4Improved
add bias tuning to lite residual modules (ablation)TinyTL-L · Carstop-1 accuracy (%): 84.2 -> 85.0Improved
add bias tuning to lite residual modules (ablation)TinyTL-L · Aircrafttop-1 accuracy (%): 76.4 -> 75.4Mixed
initialize lite residual modules with pre-trained weights instead of random (ablation)TinyTL-RandomL+B · Carstop-1 accuracy (%): 82.4 -> 85.0Improved
initialize lite residual modules with pre-trained weights instead of random (ablation)TinyTL-RandomL+B · Flowerstop-1 accuracy (%): 88.0 -> 95.5Improved
increase resolution from 224 to 320TinyTL-L+B · Carstop-1 accuracy (%): 85.0 -> 88.8Improved
use TinyTL-L+B instead of fine-tuning last layer onlyFT-Last · Carstop-1 accuracy (%): 50.9 -> 85.0Improved
use TinyTL-L+B instead of fine-tuning norm layers and last layerFT-Norm+Last · Training memorytraining memory: 192MB -> 37MBImproved
add feature extractor adaptation with last two layers tuned (ablation)TinyTL FA@352 · Carstop-1 accuracy (%): 90.7 -> 91.5Improved
train with batch size 1 instead of 8 (ablation)TinyTL · Flowers, Aircraft, Carstop-1 accuracyNo clear effect
use TinyTL instead of dynamic activation pruningActivation pruning · Flowers, Aircraft, Carsaccuracy vs memoryImproved

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