Quantization On Imagenet

Métriques

Top-1 Accuracy (%)

Résultats

Résultats de performance de divers modèles sur ce benchmark

Nom du modèle	Top-1 Accuracy (%)	Paper Title	Repository
UniQ (Ours)	71.5	Training Multi-bit Quantized and Binarized Networks with A Learnable Symmetric Quantizer
EfficientNet-B0-W4A4	76	HMQ: Hardware Friendly Mixed Precision Quantization Block for CNNs
ResNet50-W3A4	75.45	HMQ: Hardware Friendly Mixed Precision Quantization Block for CNNs
FQ-ViT (DeiT-T)	71.61	FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer
FQ-ViT (Swin-S)	82.71	FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer
FQ-ViT (ViT-B)	83.31	FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer
FQ-ViT (DeiT-B)	81.20	FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer
DenseNet-121 W8A8	73.356	HPTQ: Hardware-Friendly Post Training Quantization
ResNet-18 + PACT + R2Loss	68.45	R2 Loss: Range Restriction Loss for Model Compression and Quantization	-
MobileNetV2 W8A8	71.46	HPTQ: Hardware-Friendly Post Training Quantization
ResNet50-W4A4 (paper)	76.7	Learned Step Size Quantization
EfficientNet-B0 W8A8	74.216	HPTQ: Hardware-Friendly Post Training Quantization
FQ-ViT (Swin-T)	80.51	FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer
MPT (80) +BN	74.03	Multi-Prize Lottery Ticket Hypothesis: Finding Accurate Binary Neural Networks by Pruning A Randomly Weighted Network
ADLIK-MO-ResNet50-W4A4	77.878	Learned Step Size Quantization
FQ-ViT (ViT-L)	85.03	FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer
MobileNet-v1 + EWGS + R2Loss	69.79	R2 Loss: Range Restriction Loss for Model Compression and Quantization	-
EfficientNet-W4A4	73.8	LSQ+: Improving low-bit quantization through learnable offsets and better initialization
ADLIK-MO-ResNet50-W3A4	77.34	Learned Step Size Quantization
MixNet-W4A4	71.7	LSQ+: Improving low-bit quantization through learnable offsets and better initialization

0 of 27 row(s) selected.