dicksonyuan / caffemodelcompression

Tool to compress trained caffe weights

Caffe Model Compression

This is a python tool used to compress the trained caffe weights. For Alexnet, we got 17x compression rate (~233M bytes to 14M bytes). The idea comes from Deep Compression . This work does not implement purning and Huffman coding, but implement the Kmeans -based quantization to compress the weights of convolution and full-connected layer. One contribution of this work is using OpenMP to accelerate the Kmeans processing.

---

####Dependency

• Python/Numpy
• Caffe

####Authors

• Li, Victor
• Qiu, Junyi

####How to Build:

cd quantz_kit 
 ./build.sh

####How to use it:

caffe_model_compress: #function to compress model 
caffe_model_decompress: #function to decompress model