hi-yan / tensorflow-quantization

Experiments and tool-chains build upon TensorFlow and its quantization tools

tensorflow-quantization

Experiments and tool-chains build upon TensorFlow and its quantization tools.

Usage (Proposed)

./main.py \
  --model-name=[MODEL NAME] # The model you would like to evaluate and quantize, model names are specified later
  --quantize-bit=[NUMBER]   # Quantize the model to which bit: 8 or 16
  --model-dir=[MODEL DIR]   # Where to save/restore the model
  --train                   # Train the model or not

Models Implemented

Model files are placed under models/.

Model Name	Source File	Description
lenet	lenet_model.py	The typical LeNet implementation on MNIST dataset