AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

PYRA: Parallel Yielding Re-Activation for Training-Inference Efficient Task Adaptation

Yizhe Xiong, Hui Chen, Tianxiang Hao, Zijia Lin, Jungong Han, Yuesong Zhang, Guoxin Wang, Yongjun Bao, Guiguang Ding

ECCV 2024 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 14 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 PYRA token modulation to ToMe + LoRAToMe + LoRA · VTAB-1k high compression, ViT-B/16average accuracy: 70.43 -> 72.06Improved
add PYRA token modulation to ToMe + LoRAToMe + LoRA · VTAB-1k high compression, ViT-L/16average accuracy: 74.10 -> 75.66Improved
add PYRA token modulation at low compressionToMe + LoRA · VTAB-1k low compression, ViT-B/16average accuracy: 74.10 -> 74.69Improved
add the W_r generator (tokens) without activation (ablation)PYRA · VTAB-1k high compression, ViT-B/16average accuracy: 70.43 -> 70.54Improved
add both W_r and W_D generators without activation (ablation)PYRA · VTAB-1k high compression, ViT-B/16average accuracy: 70.43 -> 71.32Improved
add sigmoid re-activation on top of both generators (ablation)PYRA · VTAB-1k high compression, ViT-B/16average accuracy: 71.32 -> 72.06Improved
use parallel yielding generators instead of a directly learned W_{D x r} (ablation)PYRA · VTAB-1k high compression, ViT-B/16average accuracy: 71.49 -> 72.06Improved
use PYRA generators instead of a gated MLP generator (ablation)PYRA · VTAB-1k high compression, ViT-B/16average accuracy: 71.06 -> 72.06Improved
increase LoRA hidden dimension to 4x parameters instead of adding PYRA (ablation)ToMe + LoRA baseline · VTAB-1k high compression, ViT-B/16average accuracy: 70.43 -> 71.00 (PYRA: 72.06)Improved
increase modulation generator rank above 1 (ablation)PYRA · VTAB-1k high compression, ViT-B/16average accuracyNo clear effect
one-stage training with token merging instead of two-stage (ablation)ToMe + LoRA · VTAB-1k high compression, ViT-B/16average accuracy: 69.25 -> 70.43Improved
parameter-efficient distillation instead of direct PEFT on smaller backboneViT-S/16 with LoRA · VTAB-1k high compressionaverage accuracy: 71.85 -> 70.99Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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