REQuant

Requirements

This is my experiment eviroument

• python3.10
• pytorch2.3.0+cu121
• numpy 1.26.4
• torchvision 0.18.0

Usage

1. enter directory

$ cd REQuant

2. dataset

Support cifar10 and cifar100 dataset from torchvision since it's more convenient.

3. train the model

You need to specify the net you want to train using arg -net

$ python train.py -net resnet18 -dataset cifar10 -batch 128

sometimes, you might want to use warmup training by set -warm to 1 or 2, to prevent network diverge during early training phase.

The supported net args are:

inceptionv3
inceptionv4
resnet18
resnet34
resnet50
resnet101

4. test the model by uniform

$ python test_uniform.py -net resnet18 -model_path path_to_resnet18_weights_file -dataset cifar10 -bit 4 -batch 128

5. test the model by nonuniform

$ python test_nonuniform.py -net resnet18 -model_path path_to_resnet18_weights_file -dataset cifar10 -bit 4 -batch 128