Repository files navigation

import torchfrom torch.autograd import Variablefrom utee import selectormodel_raw, ds_fetcher, is_imagenet = selector.select('mnist')ds_val = ds_fetcher(batch_size=10, train=False, val=True)for idx, (data, target) in enumerate(ds_val):    data =  Variable(torch.FloatTensor(data)).cuda()    output = model_raw(data)

python3 setup.py develop --user

We provide precomputed imagenet validation dataset with 224x224x3 size. We first resize the shorter size of image to 256, then we crop 224x224 image in the center. Then we encode the cropped images to jpg string and dump to pickle.

• cd script
• Download theval224_compressed.pkl (Tsinghua /Google Drive)
• python convert.py (needs 48G memory, thanks@jnorwood )

We also provide a simple demo to quantize these models to specified bit-width with several methods, including linear method, minmax method and non-linear method.

quantize --type cifar10 --quant_method linear --param_bits 8 --fwd_bits 8 --bn_bits 8 --ngpu 1

We evaluate the performance of popular dataset and models with linear quantized method. The bit-width of running mean and running variance in BN are 10 bits for all results. (except for 32-float)

Note: ImageNet 32-float models are directly from torchvision

Here we give an overview of selected arguments ofquantize.py