Quantization on ImageNet

Metric: Weight bits (higher is better)

LeaderboardDataset

Loading chart...

Results

Sort:

#	Model↕	Weight bits▼	Augmentations	Paper	Date↕	Code
1	FQ-ViT (ViT-L)	8	No	FQ-ViT: Post-Training Quantization for Fully Qua...	2021-11-27	Code
2	FQ-ViT (ViT-B)	8	No	FQ-ViT: Post-Training Quantization for Fully Qua...	2021-11-27	Code
3	FQ-ViT (Swin-B)	8	No	FQ-ViT: Post-Training Quantization for Fully Qua...	2021-11-27	Code
4	FQ-ViT (Swin-S)	8	No	FQ-ViT: Post-Training Quantization for Fully Qua...	2021-11-27	Code
5	FQ-ViT (DeiT-B)	8	No	FQ-ViT: Post-Training Quantization for Fully Qua...	2021-11-27	Code
6	FQ-ViT (Swin-T)	8	No	FQ-ViT: Post-Training Quantization for Fully Qua...	2021-11-27	Code
7	FQ-ViT (DeiT-S)	8	No	FQ-ViT: Post-Training Quantization for Fully Qua...	2021-11-27	Code
8	Xception W8A8	8	No	HPTQ: Hardware-Friendly Post Training Quantization	2021-09-19	Code
9	EfficientNet-B0 ReLU W8A8	8	No	HPTQ: Hardware-Friendly Post Training Quantization	2021-09-19	Code
10	EfficientNet-B0-W8A8	8	No	HMQ: Hardware Friendly Mixed Precision Quantizat...	2020-07-20	Code
11	EfficientNet-B0 W8A8	8	No	HPTQ: Hardware-Friendly Post Training Quantization	2021-09-19	Code
12	DenseNet-121 W8A8	8	No	HPTQ: Hardware-Friendly Post Training Quantization	2021-09-19	Code
13	FQ-ViT (DeiT-T)	8	No	FQ-ViT: Post-Training Quantization for Fully Qua...	2021-11-27	Code
14	MobileNetV2 W8A8	8	No	HPTQ: Hardware-Friendly Post Training Quantization	2021-09-19	Code
15	ADLIK-MO-ResNet50-W4A4	4	No	Learned Step Size Quantization	2019-02-21	Code
16	ResNet50-W4A4 (paper)	4	No	Learned Step Size Quantization	2019-02-21	Code
17	EfficientNet-B0-W4A4	4	No	HMQ: Hardware Friendly Mixed Precision Quantizat...	2020-07-20	Code
18	EfficientNet-W4A4	4	No	LSQ+: Improving low-bit quantization through lea...	2020-04-20	Code
19	MixNet-W4A4	4	No	LSQ+: Improving low-bit quantization through lea...	2020-04-20	Code
20	MobileNet-v1 + EWGS + R2Loss	4	No	R2 Loss: Range Restriction Loss for Model Compre...	2023-03-14	-
21	ADLIK-MO-ResNet50-W3A4	3	No	Learned Step Size Quantization	2019-02-21	Code
22	ResNet50-W3A4	3	No	HMQ: Hardware Friendly Mixed Precision Quantizat...	2020-07-20	Code
23	ResNet-18 + PACT + R2Loss	2	No	R2 Loss: Range Restriction Loss for Model Compre...	2023-03-14	-

#1FQ-ViT (ViT-L)
8
Weight bits· 2021-11-27
FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer Code
#2FQ-ViT (ViT-B)
8
Weight bits· 2021-11-27
FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer Code
#3FQ-ViT (Swin-B)
8
Weight bits· 2021-11-27
FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer Code
#4FQ-ViT (Swin-S)
8
Weight bits· 2021-11-27
FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer Code
#5FQ-ViT (DeiT-B)
8
Weight bits· 2021-11-27
FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer Code
#6FQ-ViT (Swin-T)
8
Weight bits· 2021-11-27
FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer Code
#7FQ-ViT (DeiT-S)
8
Weight bits· 2021-11-27
FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer Code
#8Xception W8A8
8
Weight bits· 2021-09-19
HPTQ: Hardware-Friendly Post Training Quantization Code
#9EfficientNet-B0 ReLU W8A8
8
Weight bits· 2021-09-19
HPTQ: Hardware-Friendly Post Training Quantization Code
#10EfficientNet-B0-W8A8SOTA
8
Weight bits· 2020-07-20
HMQ: Hardware Friendly Mixed Precision Quantization Block for CNNs Code
#11EfficientNet-B0 W8A8
8
Weight bits· 2021-09-19
HPTQ: Hardware-Friendly Post Training Quantization Code
#12DenseNet-121 W8A8
8
Weight bits· 2021-09-19
HPTQ: Hardware-Friendly Post Training Quantization Code
#13FQ-ViT (DeiT-T)
8
Weight bits· 2021-11-27
FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer Code
#14MobileNetV2 W8A8
8
Weight bits· 2021-09-19
HPTQ: Hardware-Friendly Post Training Quantization Code
#15ADLIK-MO-ResNet50-W4A4SOTA
4
Weight bits· 2019-02-21
Learned Step Size Quantization Code
#16ResNet50-W4A4 (paper)
4
Weight bits· 2019-02-21
Learned Step Size Quantization Code
#17EfficientNet-B0-W4A4
4
Weight bits· 2020-07-20
HMQ: Hardware Friendly Mixed Precision Quantization Block for CNNs Code
#18EfficientNet-W4A4
4
Weight bits· 2020-04-20
LSQ+: Improving low-bit quantization through learnable offsets and better initialization Code
#19MixNet-W4A4
4
Weight bits· 2020-04-20
LSQ+: Improving low-bit quantization through learnable offsets and better initialization Code
#20MobileNet-v1 + EWGS + R2Loss
4
Weight bits· 2023-03-14
R2 Loss: Range Restriction Loss for Model Compression and Quantization
#21ADLIK-MO-ResNet50-W3A4
3
Weight bits· 2019-02-21
Learned Step Size Quantization Code
#22ResNet50-W3A4
3
Weight bits· 2020-07-20
HMQ: Hardware Friendly Mixed Precision Quantization Block for CNNs Code
#23ResNet-18 + PACT + R2Loss
2
Weight bits· 2023-03-14
R2 Loss: Range Restriction Loss for Model Compression and Quantization