基于pytorch的三种模型定点化方法（dynamic qu，Post-training static quantization，Quantization-aware training (QAT) ）

%matplotlib inline

(experimental) Static Quantization with Eager Mode in PyTorch

Author: Raghuraman Krishnamoorthi <https://github.com/raghuramank100>_

Edited by: Seth Weidman <https://github.com/SethHWeidman/>_

This tutorial shows how to do post-training static quantization, as well as illustrating
two more advanced techniques - per-channel quantization and quantization-aware training -
to further improve the model’s accuracy. Note that quantization is currently only supported
for CPUs, so we will not be utilizing GPUs / CUDA in this tutorial.

By the end of this tutorial, you will see how quantization in PyTorch can result in
significant decreases in model size while increasing speed. Furthermore, you’ll see how
to easily apply some advanced quantization techniques shown
here <https://arxiv.org/abs/1806.08342>_ so that your quantized models take much less
of an accuracy hit than they would otherwise.

Warning: we use a lot of boilerplate code from other PyTorch repos to, for example,
define the MobileNetV2 model archtecture, define data loaders, and so on. We of course
encourage you to read it; but if you want to get to the quantization features, feel free
to skip to the “4. Post-training static quantization” section.

We’ll start by doing the necessary imports:

import numpy as np
import torch
import torch.nn as nn
import torchvision
from torch.utils.data import DataLoader
from torchvision import datasets
import torchvision.transforms as transforms
import os
import time
import sys
import torch.quantization

# # Setup warnings
import warnings
warnings.filterwarnings(
    action='ignore',
    category=DeprecationWarning,
    module=r'.*'
)
warnings.filterwarnings(
    action='default',
    module=r'torch.quantization'
)

# Specify random seed for repeatable results
torch.manual_seed(191009)

<torch._C.Generator at 0x7feab6b90110>

1. Model architecture

---

We first define the MobileNetV2 model architecture, with several notable modifications
to enable quantization:

• Replacing addition with nn.quantized.FloatFunctional
• Insert QuantStub and DeQuantStub at the beginning and end of the network.
• Replace ReLU6 with ReLU

Note: this code is taken from
here <https://github.com/pytorch/vision/blob/master/torchvision/models/mobilenet.py>_.

from torch.quantization import QuantStub, DeQuantStub

def _make_divisible(v, divisor, min_value=None):
    """
    This function is taken from the original tf repo.
    It ensures that all layers have a channel number that is divisible by 8
    It can be seen here:
    https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py
    :param v:
    :param divisor:
    :param min_value:
    :return:
    """
    if min_value is None:
        min_value = divisor
    new_v = max(min_value, int(v + divisor / 2) // divisor * divisor)
    # Make sure that round down does not go down by more than 10%.
    if new_v < 0.9 * v:
        new_v += divisor
    return new_v


class ConvBNReLU(nn.Sequential):
    def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, groups=1):
        padding = (kernel_size - 1) // 2
        super(ConvBNReLU, self).__init__(
            nn.Conv2d(in_planes, out_planes, kernel_size, stride, padding, groups=groups, bias=False),
            nn.BatchNorm2d(out_planes, momentum=0.1),
            # Replace with ReLU
            nn.ReLU(inplace=False)
        )


class InvertedResidual(nn.Module):
    def __init__(self, inp, oup, stride, expand_ratio):
        super(InvertedResidual, self).__init__()
        self.stride = stride
        assert stride in [1, 2]

        hidden_dim = int(round(inp * expand_ratio))
        self.use_res_connect = self.stride == 1 and inp == oup

        layers = []
        if expand_ratio != 1:
            # pw
            layers.append(ConvBNReLU(inp, hidden_dim, kernel_size=1))
        layers.extend([
            # dw
            ConvBNReLU(hidden_dim, hidden_dim, stride=stride, groups=hidden_dim),
            # pw-linear
            nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False),
            nn.BatchNorm2d(oup, momentum=0.1),
        ])
        self.conv = nn.Sequential(*layers)
        # Replace torch.add with floatfunctional
        self.skip_add = nn.quantized.FloatFunctional()

    def forward(self, x):
        if self.use_res_connect:
            return self.skip_add.add(x, self.conv(x))
        else:
            return self.conv(x)


class MobileNetV2(nn.Module):
    def __init__(self, num_classes=1000, width_mult=1.0, inverted_residual_setting=None, round_nearest=8):
        """
        MobileNet V2 main class

        Args:
            num_classes (int): Number of classes
            width_mult (float): Width multiplier - adjusts number of channels in each layer by this amount
            inverted_residual_setting: Network structure
            round_nearest (int): Round the number of channels in each layer to be a multiple of this number
            Set to 1 to turn off rounding
        """
        super(MobileNetV2, self).__init__()
        block = InvertedResidual
        input_channel = 32
        last_channel = 1280

        if inverted_residual_setting is None:
            inverted_residual_setting = [
                # t, c, n, s
                [1, 16, 1, 1],
                [6, 24, 2, 2],
                [6, 32, 3, 2],
                [6, 64, 4, 2],
                [6, 96, 3, 1],
                [6, 160, 3<