Awesome Quantization Paper lists with Codes.
- Quantized weights + activation, during training and test (Training aware Quantization)
- LSQ
- QIL
- LQ-Nets
- Quantized weights + activation, during test (Post training Quantization)
- ZeroQ
- OCS
- ACIQ
Quantize method | Tiitle | conference | code | Train / Inference |
---|---|---|---|---|
LSQ | Learned Step Size Quantization | ICLR 2020 | Unofficial, Pytorch | Training aware |
QIL | Learning to Quantize Deep Networks by Optimizing Quantization Intervals with Task Loss | CVPR 2019 | Unofficial, Pytorch | Training aware |
LQ-Nets | Learned Quantization for Highly Accurate and Compact Deep Neural Networks | ECCV 2018 | Official, Tensorflow | Training aware |
PACT | PACT: Parameterized Clipping Activation for Quantized Neural Networks | Training aware | ||
ZeroQ | ZeroQ: A Novel Zero Shot Quantization Framework | CVPR 2020 | Official, Pytorch | Post training |
OCS | Improving Neural Network Quantization without Retraining using Outlier Channel Splitting | ICML 2019 | Official, Pytorch | Post training |
ACIQ | Post-training 4-bit quantization of convolution networks for rapid-deployment | NIPS 2019 | Official, Pytorch | Post training |
- Code reference: pytorch-quant [github, Pytorch]
Quantize method | Tiitle | conference | Train / Inference | scale |
---|---|---|---|---|
Minmax | Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference | CVPR 2018 | Training aware | minmax |
Log Minmax | Convolutional Neural Networks using Logarithmic Data Representation | arXiv 2016 | Training aware | log minmax |
Tanh | Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activations | arXiv 2016 | Training aware | tanh |
- Deep Compression
- DoReFa-Net
- BNN
- XNOR-Net
- Awesome Deep Neural Network Compression [github]
# Author
- Subin Yang
- ysb8049@gmail.com