AI Paper Atlas

Research map / Architectures and efficiency / Quantization

RMSMP: A Novel Deep Neural Network Quantization Framework With Row-Wise Mixed Schemes and Multiple Precisions

Sung-En Chang, Yanyu Li, Mengshu Sun, Weiwen Jiang, Sijia Liu, Yanzhi Wang, Xue Lin

ICCV 2021 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 8 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 5% Fixed-W8A4 filters (row-wise multi-precision) to Fixed-W4A4 (ablation)Fixed-W4A4 · ImageNet, ResNet-18top-1 accuracy (%): 68.66 -> 70.48Improved
add PoT-W4A4 filters to Fixed-W4A4 + Fixed-W8A4 (RMSMP 65:30:5) (ablation)Fixed-W4A4 + Fixed-W8A4 · ImageNet, ResNet-18top-1 accuracy (%): 70.48 -> 70.73Improved
add PoT-W4A4 filters to Fixed-W4A4 + Fixed-W8A4 (RMSMP) (ablation)Fixed-W4A4 + Fixed-W8A4 · ImageNet, ResNet-50top-1 accuracy (%): 76.68 -> 76.62No clear effect
add PoT-W4A4 filters to Fixed-W4A4 + Fixed-W8A4 (RMSMP) (ablation)Fixed-W4A4 + Fixed-W8A4 · ImageNet, MobileNet-v2top-1 accuracy (%): 68.98 -> 69.02No clear effect
mix PoT-W4A4 with Fixed-W4A4 filters (1:1) instead of Fixed alone (ablation)Fixed-W4A4 · ImageNet, ResNet-18top-1 accuracy (%): 68.66 -> 67.98Hurt
quantize first and last layers to 4-bit instead of 8-bit Fixed (ablation)Fixed-W4A4 · ImageNet, ResNet-18top-1 accuracy (%): 69.72 -> 68.66Hurt
quantize first and last layers to 4-bit instead of 8-bit Fixed (ablation)Fixed-W4A4 · FPGA XC7Z020, ResNet-18latency (ms): 122.6 -> 99.3Improved
use RMSMP-1 (60:35:5) instead of Fixed-only quantizationFixed single scheme · FPGA XC7Z020, ResNet-18latency (ms): 122.6 -> 40.7Improved
use RMSMP-2 (65:30:5) instead of Fixed-only quantizationFixed single scheme · FPGA XC7Z045, ResNet-18latency (ms): 31.4 -> 8.6Improved
use RMSMP instead of MSQMSQ · ImageNet, ResNet-18top-1 accuracy (%): 70.27 -> 70.73Improved
use RMSMP instead of MSQ-2 on XC7Z045MSQ-2 · FPGA XC7Z045, ResNet-18latency (ms): 10.1 -> 8.6Improved
quantize BERT with RMSMP rather than 32-bitBERT · SST-2accuracy (%): 93.00 -> 92.87No 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