Merge branch 'master' of http://62.234.201.16/hao/Model-Transfer-Adaptability

ykl/AlexNet/README.md

+# AlexNet 量化说明
+
+## ptq部分
+
++ INT/POT/FLOAT量化都采用相同的框架，可以通过`quant_type`进行确定
+
++ 量化范围：均采用有符号对称量化，且将zeropoint定为0
+
++ 量化策略：在第一次forward执行伪量化，统计每层的x和weight范围；后续interface在卷积/池化层中使用量化后的值进行运算。量化均通过放缩至对应范围后从量化点列表取最近点进行实现。
+
++ bias说明：每种量化模式下，bias采用相同的量化策略(INT:32bit量化，POT:8bit量化，FP8:FP16-E7量化)。bias量化损失对结果影响小，该量化策略也对运算硬件实现影响不大，但是在代码实现上可以更加高效，故采用。（英伟达的量化策略甚至直接舍弃了bias）  
+
+  + 关于量化策略设置，可以更改`module.py`中的`bias_qmax`函数和`utils.py`中的`build_bias_list`函数
+  + 由于INT量化位宽较高，使用量化表开销过大，直接使用round_操作即可
+
++ 量化点选择：
+
+  + INT：取INT2-INT16（INT16后相比全精度无损失）
+  + POT：取POT2-POT8  （POT8之后容易出现Overflow）
+  + FP8：取E1-E6 （E0相当于INT量化，E7相当于POT量化，直接取相应策略效果更好）
+  + 支持调整FP的位宽
+  + 关于量化点选择，可以更改`utils.py`中的`bit_list`函数
+
++ 量化结果：
+
+  FP32-acc：85.08
+
+  <img src="image/table.png" alt="table" style="zoom: 33%;" />
+
++ 数据拟合：
+
+  matlab导入数据，选择列向量
+
+  + js_flops - acc_loss
+
+    Rational: Numerator degree 2 / Denominator degree 2
+
+    - [ ] center and scale
+
+    ![fig1](image/fig1.png)
+
+    - [x] center and scale
+
+    ![fig2](image/fig2.png)
+
+  + js_param - acc_loss
+
+    Rational: Numerator degree 2 / Denominator degree 2
+
+    - [ ] center and scale
+
+    ![fig3](image/fig3.png)
+
+    - [x] center and scale
+
+    ![fig4](image/fig4.png)
\ No newline at end of file

ykl/AlexNet/extract_ratio.py

+import sys
+import os
+
+
+# 从get_param.py输出重定向文件val.txt中提取参数量和计算量
+def extract_ratio():
+    fr = open('param_flops.txt','r')
+    lines = fr.readlines()
+    layer = []
+    par_ratio = []
+    flop_ratio = []
+    for line in lines:
+        if '(' in line and ')' in line:
+            layer.append(line.split(')')[0].split('(')[1])
+            r1 = line.split('%')[0].split(',')[-1]
+            r1 = float(r1)
+            par_ratio.append(r1)
+            r2 = line.split('%')[-2].split(',')[-1]
+            r2 = float(r2)
+            flop_ratio.append(r2)
+
+    return layer, par_ratio, flop_ratio
+
+
+if __name__ == "__main__":
+    layer, par_ratio, flop_ratio = extract_ratio()
+    print(layer)
+    print(par_ratio)
+    print(flop_ratio)
\ No newline at end of file

ykl/AlexNet/function.py

+from torch.autograd import Function
+
+
+class FakeQuantize(Function):
+
+    @staticmethod
+    def forward(ctx, x, qparam):
+        x = qparam.quantize_tensor(x)
+        x = qparam.dequantize_tensor(x)
+        return x
+
+    @staticmethod
+    def backward(ctx, grad_output):
+        return grad_output, None
\ No newline at end of file

ykl/AlexNet/get_param_flops.py

+from model import *
+
+import torch
+from ptflops import get_model_complexity_info
+
+
+if __name__ == "__main__":
+    model = AlexNet()
+    full_file = 'ckpt/cifar10_AlexNet.pt'
+    model.load_state_dict(torch.load(full_file))
+
+    flops, params = get_model_complexity_info(model, (3, 32, 32), as_strings=True, print_per_layer_stat=True)
+

ykl/AlexNet/gol.py

+# -*- coding: utf-8 -*-
+
+# 用于多个module之间共享全局变量
+def _init():  # 初始化
+    global _global_dict
+    _global_dict = {}
+ 
+def set_value(value,is_bias=False):
+    # 定义一个全局变量
+    if is_bias:
+        _global_dict[0] = value
+    else:
+        _global_dict[1] = value
+ 
+ 
+def get_value(is_bias=False): # 给bias独立于各变量外的精度
+    if is_bias:
+        return _global_dict[0]
+    else:
+        return _global_dict[1]  
+

ykl/AlexNet/model.py

+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from module import *
+import module
+
+
+class AlexNet(nn.Module):
+
+    def __init__(self, num_channels=3, num_classes=10):
+        super(AlexNet, self).__init__()
+
+        # original size 32x32
+        self.conv1 = nn.Conv2d(num_channels, 32, kernel_size=3, padding=1, bias=True)
+        self.relu1 = nn.ReLU(inplace=True)
+        self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2)                  # output[48, 27, 27]
+        self.conv2 = nn.Conv2d(32, 64, kernel_size=3, padding=1, bias=True)           # output[128, 27, 27]
+        self.relu2 = nn.ReLU(inplace=True)
+        self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2)                  # output[128, 13, 13]
+        self.conv3 = nn.Conv2d(64, 128, kernel_size=3, padding=1, bias=True)          # output[192, 13, 13]
+        self.relu3 = nn.ReLU(inplace=True)
+        self.conv4 = nn.Conv2d(128, 256, kernel_size=3, padding=1, bias=True)          # output[192, 13, 13]
+        self.relu4 = nn.ReLU(inplace=True)
+        self.conv5 = nn.Conv2d(256, 256, kernel_size=3, padding=1, bias=True)          # output[128, 13, 13]
+        self.relu5 = nn.ReLU(inplace=True)
+        self.pool5 = nn.MaxPool2d(kernel_size=3, stride=2)
+
+        self.drop1 = nn.Dropout(p=0.5)
+        self.fc1 = nn.Linear(256 * 3 * 3, 1024, bias=True)
+        self.relu6 = nn.ReLU(inplace=True)
+        self.drop2 = nn.Dropout(p=0.5)
+        self.fc2 = nn.Linear(1024, 512, bias=True)
+        self.relu7 = nn.ReLU(inplace=True)
+        self.fc3 = nn.Linear(512, num_classes, bias=True)
+
+    def forward(self, x):
+        x = self.conv1(x)
+        x = self.relu1(x)
+        x = self.pool1(x)
+        x = self.conv2(x)
+        x = self.relu2(x)
+        x = self.pool2(x)
+        x = self.conv3(x)
+        x = self.relu3(x)
+        x = self.conv4(x)
+        x = self.relu4(x)
+        x = self.conv5(x)
+        x = self.relu5(x)
+        x = self.pool5(x)
+
+        x = torch.flatten(x, start_dim=1)
+
+        x = self.drop1(x)
+        x = self.fc1(x)
+        x = self.relu6(x)
+        x = self.drop2(x)
+        x = self.fc2(x)
+        x = self.relu7(x)
+        x = self.fc3(x)
+        return x
+
+    def quantize(self, quant_type, num_bits=8, e_bits=3): 
+        # e_bits仅当使用FLOAT量化时用到
+
+        self.qconv1 = QConv2d(quant_type, self.conv1, qi=True, qo=True, num_bits=num_bits, e_bits=e_bits)
+        self.qrelu1 = QReLU(quant_type, num_bits=num_bits, e_bits=e_bits)
+        self.qpool1 = QMaxPooling2d(quant_type, kernel_size=2, stride=2, padding=0, num_bits=num_bits, e_bits=e_bits)
+        self.qconv2 = QConv2d(quant_type, self.conv2, qi=False, qo=True, num_bits=num_bits, e_bits=e_bits)
+        self.qrelu2 = QReLU(quant_type, num_bits=num_bits, e_bits=e_bits)
+        self.qpool2 = QMaxPooling2d(quant_type, kernel_size=2, stride=2, padding=0, num_bits=num_bits, e_bits=e_bits)
+        self.qconv3 = QConv2d(quant_type, self.conv3, qi=False, qo=True, num_bits=num_bits, e_bits=e_bits)
+        self.qrelu3 = QReLU(quant_type, num_bits=num_bits, e_bits=e_bits)
+        self.qconv4 = QConv2d(quant_type, self.conv4, qi=False, qo=True, num_bits=num_bits, e_bits=e_bits)
+        self.qrelu4 = QReLU(quant_type, num_bits=num_bits, e_bits=e_bits)
+        self.qconv5 = QConv2d(quant_type, self.conv5, qi=False, qo=True, num_bits=num_bits, e_bits=e_bits)
+        self.qrelu5 = QReLU(quant_type, num_bits=num_bits, e_bits=e_bits)
+        self.qpool5 = QMaxPooling2d(quant_type, kernel_size=3, stride=2, padding=0, num_bits=num_bits, e_bits=e_bits)
+
+        self.qfc1 = QLinear(quant_type, self.fc1, qi=False, qo=True, num_bits=num_bits, e_bits=e_bits)
+        self.qrelu6 = QReLU(quant_type, num_bits=num_bits, e_bits=e_bits)
+        self.qfc2 = QLinear(quant_type, self.fc2, qi=False, qo=True, num_bits=num_bits, e_bits=e_bits)
+        self.qrelu7 = QReLU(quant_type, num_bits=num_bits, e_bits=e_bits)
+        self.qfc3 = QLinear(quant_type, self.fc3, qi=False, qo=True, num_bits=num_bits, e_bits=e_bits)
+
+    def quantize_forward(self, x):
+        x = self.qconv1(x)
+        x = self.qrelu1(x)
+        x = self.qpool1(x)
+        x = self.qconv2(x)
+        x = self.qrelu2(x)
+        x = self.qpool2(x)
+        x = self.qconv3(x)
+        x = self.qrelu3(x)
+        x = self.qconv4(x)
+        x = self.qrelu4(x)
+        x = self.qconv5(x)
+        x = self.qrelu5(x)
+        x = self.qpool5(x)
+
+        x = torch.flatten(x, start_dim=1)
+
+        x = self.drop1(x)
+        x = self.qfc1(x)
+        x = self.qrelu6(x)
+        x = self.drop2(x)
+        x = self.qfc2(x)
+        x = self.qrelu7(x)
+        x = self.qfc3(x)
+        return x
+
+    def freeze(self):
+        self.qconv1.freeze()
+        self.qrelu1.freeze(self.qconv1.qo)
+        self.qpool1.freeze(self.qconv1.qo)
+        self.qconv2.freeze(self.qconv1.qo)
+        self.qrelu2.freeze(self.qconv2.qo)
+        self.qpool2.freeze(self.qconv2.qo)
+        self.qconv3.freeze(self.qconv2.qo)
+        self.qrelu3.freeze(self.qconv3.qo)
+        self.qconv4.freeze(self.qconv3.qo)
+        self.qrelu4.freeze(self.qconv4.qo)
+        self.qconv5.freeze(self.qconv4.qo)
+        self.qrelu5.freeze(self.qconv5.qo)
+        self.qpool5.freeze(self.qconv5.qo)
+        self.qfc1.freeze(self.qconv5.qo)
+        self.qrelu6.freeze(self.qfc1.qo)
+        self.qfc2.freeze(self.qfc1.qo)
+        self.qrelu7.freeze(self.qfc2.qo)
+        self.qfc3.freeze(self.qfc2.qo)
+
+    def quantize_inference(self, x):
+        qx = self.qconv1.qi.quantize_tensor(x)
+        qx = self.qconv1.quantize_inference(qx)
+        qx = self.qrelu1.quantize_inference(qx)
+        qx = self.qpool1.quantize_inference(qx)
+        qx = self.qconv2.quantize_inference(qx)
+        qx = self.qrelu2.quantize_inference(qx)
+        qx = self.qpool2.quantize_inference(qx)
+        qx = self.qconv3.quantize_inference(qx)
+        qx = self.qrelu3.quantize_inference(qx)
+        qx = self.qconv4.quantize_inference(qx)
+        qx = self.qrelu4.quantize_inference(qx)
+        qx = self.qconv5.quantize_inference(qx)
+        qx = self.qrelu5.quantize_inference(qx)
+        qx = self.qpool5.quantize_inference(qx)
+
+        qx = torch.flatten(qx, start_dim=1)
+
+        qx = self.qfc1.quantize_inference(qx)
+        qx = self.qrelu6.quantize_inference(qx)
+        qx = self.qfc2.quantize_inference(qx)
+        qx = self.qrelu7.quantize_inference(qx)
+        qx = self.qfc3.quantize_inference(qx)
+        out = self.qfc3.qo.dequantize_tensor(qx)
+        return out

ykl/AlexNet/module.py

+import math
+import numpy as np
+import gol
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch.autograd import Variable
+
+from function import FakeQuantize
+
+# 获取最近的量化值
+def get_nearest_val(quant_type,x,is_bias=False):
+    if quant_type=='INT':
+        return x.round_()
+    
+    plist = gol.get_value(is_bias)
+    # print('get')
+    # print(plist)
+    shape = x.shape
+    xhard = x.view(-1)
+    plist = plist.type_as(x)
+    # 取最近幂次作为索引
+    idx = (xhard.unsqueeze(0) - plist.unsqueeze(1)).abs().min(dim=0)[1]
+    xhard = plist[idx].view(shape)
+    xout = (xhard - x).detach() + x
+    return xout
+
+# 采用对称有符号量化时，获取量化范围最大值
+def get_qmax(quant_type,num_bits=None, e_bits=None):
+    if quant_type == 'INT':
+        qmax = 2. ** (num_bits - 1) - 1
+    elif quant_type == 'POT':
+        qmax = 1
+    else: #FLOAT
+        m_bits = num_bits - 1 - e_bits
+        dist_m = 2 ** (-m_bits)
+        e = 2 ** (e_bits - 1)
+        expo = 2 ** e
+        m = 2 ** m_bits -1
+        frac = 1. + m * dist_m
+        qmax = frac * expo
+    return qmax
+
+# 都采用有符号量化，zeropoint都置为0
+def calcScaleZeroPoint(min_val, max_val, qmax):
+    scale = torch.max(max_val.abs(),min_val.abs()) / qmax
+    zero_point = torch.tensor(0.)
+    return scale, zero_point
+
+# 将输入进行量化，输入输出都为tensor
+def quantize_tensor(quant_type, x, scale, zero_point, qmax, is_bias=False):
+    # 量化后范围，直接根据位宽确定
+    qmin = -qmax
+    q_x = zero_point + x / scale
+
+    q_x.clamp_(qmin, qmax)
+    q_x = get_nearest_val(quant_type, q_x, is_bias)
+    return q_x
+
+# bias使用不同精度，需要根据量化类型指定num_bits/e_bits
+def bias_qmax(quant_type):
+    if quant_type == 'INT':
+        return get_qmax(quant_type, 64)
+    elif quant_type == 'POT':
+        return get_qmax(quant_type)
+    else:
+        return get_qmax(quant_type, 16, 5)
+        
+
+# 转化为FP32，不需再做限制
+def dequantize_tensor(q_x, scale, zero_point):
+    return scale * (q_x - zero_point)
+
+
+class QParam(nn.Module):
+
+    def __init__(self,quant_type, num_bits=8, e_bits=3):
+        super(QParam, self).__init__()
+        self.quant_type = quant_type
+        self.num_bits = num_bits
+        self.e_bits = e_bits
+        self.qmax = get_qmax(quant_type, num_bits, e_bits)
+
+        scale = torch.tensor([], requires_grad=False)
+        zero_point = torch.tensor([], requires_grad=False)
+        min = torch.tensor([], requires_grad=False)
+        max = torch.tensor([], requires_grad=False)
+        # 通过注册为register，使得buffer可以被记录到state_dict
+        self.register_buffer('scale', scale)
+        self.register_buffer('zero_point', zero_point)
+        self.register_buffer('min', min)
+        self.register_buffer('max', max)
+
+    # 更新统计范围及量化参数
+    def update(self, tensor):
+        if self.max.nelement() == 0 or self.max.data < tensor.max().data:
+            self.max.data = tensor.max().data
+        self.max.clamp_(min=0)
+
+        if self.min.nelement() == 0 or self.min.data > tensor.min().data:
+            self.min.data = tensor.min().data
+        self.min.clamp_(max=0)
+
+        self.scale, self.zero_point = calcScaleZeroPoint(self.min, self.max, self.qmax)
+
+    def quantize_tensor(self, tensor):
+        return quantize_tensor(self.quant_type, tensor, self.scale, self.zero_point, self.qmax)
+
+    def dequantize_tensor(self, q_x):
+        return dequantize_tensor(q_x, self.scale, self.zero_point)
+
+    # 该方法保证了可以从state_dict里恢复
+    def _load_from_state_dict(self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys,
+                              error_msgs):
+        key_names = ['scale', 'zero_point', 'min', 'max']
+        for key in key_names:
+            value = getattr(self, key)
+            value.data = state_dict[prefix + key].data
+            state_dict.pop(prefix + key)
+
+    # 该方法返回值将是打印该对象的结果
+    def __str__(self):
+        info = 'scale: %.10f ' % self.scale
+        info += 'zp: %.6f ' % self.zero_point
+        info += 'min: %.6f ' % self.min
+        info += 'max: %.6f' % self.max
+        return info
+
+
+# 作为具体量化层的父类，qi和qo分别为量化输入/输出
+class QModule(nn.Module):
+
+    def __init__(self,quant_type, qi=True, qo=True, num_bits=8, e_bits=3):
+        super(QModule, self).__init__()
+        if qi:
+            self.qi = QParam(quant_type,num_bits, e_bits)
+        if qo:
+            self.qo = QParam(quant_type,num_bits, e_bits)
+        self.quant_type = quant_type
+        self.num_bits = num_bits
+        self.e_bits = e_bits
+        self.bias_qmax = bias_qmax(quant_type)
+
+    def freeze(self):
+        pass  # 空语句
+
+    def quantize_inference(self, x):
+        raise NotImplementedError('quantize_inference should be implemented.')
+
+
+"""
+QModule  量化卷积
+
+:quant_type: 量化类型
+:conv_module: 卷积模块
+:qi: 是否量化输入特征图
+:qo: 是否量化输出特征图
+:num_bits: 8位bit数
+"""
+
+
+class QConv2d(QModule):
+
+    def __init__(self, quant_type, conv_module, qi=True, qo=True, num_bits=8, e_bits=3):
+        super(QConv2d, self).__init__(quant_type, qi, qo, num_bits, e_bits)
+        self.conv_module = conv_module
+        self.qw = QParam(quant_type, num_bits,e_bits)
+        self.register_buffer('M', torch.tensor([], requires_grad=False))  # 将M注册为buffer
+
+    # freeze方法可以固定真量化的权重参数，并将该值更新到原全精度层上，便于散度计算
+    def freeze(self, qi=None, qo=None):
+
+        if hasattr(self, 'qi') and qi is not None:
+            raise ValueError('qi has been provided in init function.')
+        if not hasattr(self, 'qi') and qi is None:
+            raise ValueError('qi is not existed, should be provided.')
+
+        if hasattr(self, 'qo') and qo is not None:
+            raise ValueError('qo has been provided in init function.')
+        if not hasattr(self, 'qo') and qo is None:
+            raise ValueError('qo is not existed, should be provided.')
+
+        # 这里因为在池化或者激活的输入，不需要对最大值和最小是进行额外的统计，会共享相同的输出
+        if qi is not None:
+            self.qi = qi
+        if qo is not None:
+            self.qo = qo
+
+        # 根据https://zhuanlan.zhihu.com/p/156835141, 这是式3 的系数
+        self.M.data = (self.qw.scale * self.qi.scale / self.qo.scale).data
+
+        self.conv_module.weight.data = self.qw.quantize_tensor(self.conv_module.weight.data)
+        self.conv_module.weight.data = self.conv_module.weight.data - self.qw.zero_point
+
+        self.conv_module.bias.data = quantize_tensor(self.quant_type,
+                                                    self.conv_module.bias.data, scale=self.qi.scale * self.qw.scale,
+                                                    zero_point=0.,qmax=self.bias_qmax, is_bias=True)
+        
+    def forward(self, x):  # 前向传播,输入张量,x为浮点型数据
+        if hasattr(self, 'qi'):
+            self.qi.update(x)
+            x = FakeQuantize.apply(x, self.qi)  # 对输入张量X完成量化
+
+        # foward前更新qw，保证量化weight时候scale正确
+        self.qw.update(self.conv_module.weight.data)
+        # 注意:此处主要为了统计各层x和weight范围，未对bias进行量化操作
+        tmp_wgt = FakeQuantize.apply(self.conv_module.weight, self.qw)
+        x = F.conv2d(x, tmp_wgt, self.conv_module.bias,
+                     stride=self.conv_module.stride,
+                     padding=self.conv_module.padding, dilation=self.conv_module.dilation,
+                     groups=self.conv_module.groups)
+
+        if hasattr(self, 'qo'):
+            self.qo.update(x)
+            x = FakeQuantize.apply(x, self.qo)
+
+        return x
+
+    # 利用公式 q_a = M(\sigma(q_w-Z_w)(q_x-Z_x) + q_b)
+    def quantize_inference(self, x):  # 此处input为已经量化的qx
+        x = x - self.qi.zero_point
+        x = self.conv_module(x)
+        x = self.M * x
+
+        x = get_nearest_val(self.quant_type,x)
+        x = x + self.qo.zero_point
+        return x
+
+
+class QLinear(QModule):
+
+    def __init__(self, quant_type, fc_module, qi=True, qo=True, num_bits=8, e_bits=3):
+        super(QLinear, self).__init__(quant_type, qi, qo, num_bits, e_bits)
+        self.fc_module = fc_module
+        self.qw = QParam(quant_type, num_bits, e_bits)
+        self.register_buffer('M', torch.tensor([], requires_grad=False))  # 将M注册为buffer
+
+    def freeze(self, qi=None, qo=None):
+
+        if hasattr(self, 'qi') and qi is not None:
+            raise ValueError('qi has been provided in init function.')
+        if not hasattr(self, 'qi') and qi is None:
+            raise ValueError('qi is not existed, should be provided.')
+
+        if hasattr(self, 'qo') and qo is not None:
+            raise ValueError('qo has been provided in init function.')
+        if not hasattr(self, 'qo') and qo is None:
+            raise ValueError('qo is not existed, should be provided.')
+
+        if qi is not None:
+            self.qi = qi
+        if qo is not None:
+            self.qo = qo
+        
+        self.M.data = (self.qw.scale * self.qi.scale / self.qo.scale).data
+
+        self.fc_module.weight.data = self.qw.quantize_tensor(self.fc_module.weight.data)
+        self.fc_module.weight.data = self.fc_module.weight.data - self.qw.zero_point
+        self.fc_module.bias.data = quantize_tensor(self.quant_type,
+                                                    self.fc_module.bias.data, scale=self.qi.scale * self.qw.scale,
+                                                    zero_point=0., qmax=self.bias_qmax, is_bias=True)
+
+
+    def forward(self, x):
+        if hasattr(self, 'qi'):
+            self.qi.update(x)
+            x = FakeQuantize.apply(x, self.qi)
+
+        self.qw.update(self.fc_module.weight.data)
+        tmp_wgt = FakeQuantize.apply(self.fc_module.weight, self.qw)
+        x = F.linear(x, tmp_wgt, self.fc_module.bias)
+
+        if hasattr(self, 'qo'):
+            self.qo.update(x)
+            x = FakeQuantize.apply(x, self.qo)
+
+        return x
+
+    def quantize_inference(self, x):
+        x = x - self.qi.zero_point
+        x = self.fc_module(x)
+        x = self.M * x
+        x = get_nearest_val(self.quant_type,x)
+        x = x + self.qo.zero_point
+
+        return x
+
+
+class QReLU(QModule):
+
+    def __init__(self,quant_type, qi=False, qo=True, num_bits=8, e_bits=3):
+        super(QReLU, self).__init__(quant_type, qi, qo, num_bits, e_bits)
+
+    def freeze(self, qi=None):
+
+        if hasattr(self, 'qi') and qi is not None:
+            raise ValueError('qi has been provided in init function.')
+        if not hasattr(self, 'qi') and qi is None:
+            raise ValueError('qi is not existed, should be provided.')
+
+        if qi is not None:
+            self.qi = qi
+
+    def forward(self, x):
+        if hasattr(self, 'qi'):
+            self.qi.update(x)
+            x = FakeQuantize.apply(x, self.qi)
+
+        x = F.relu(x)
+
+        return x
+
+    def quantize_inference(self, x):
+        x = x.clone()
+        x[x < self.qi.zero_point] = self.qi.zero_point
+        return x
+
+
+class QMaxPooling2d(QModule):
+
+    def __init__(self, quant_type, kernel_size=3, stride=1, padding=0, qi=False, qo=True, num_bits=8,e_bits=3):
+        super(QMaxPooling2d, self).__init__(quant_type, qi, qo, num_bits, e_bits)
+        self.kernel_size = kernel_size
+        self.stride = stride
+        self.padding = padding
+        
+    def freeze(self, qi=None):
+        if hasattr(self, 'qi') and qi is not None:
+            raise ValueError('qi has been provided in init function.')
+        if not hasattr(self, 'qi') and qi is None:
+            raise ValueError('qi is not existed, should be provided.')
+        if qi is not None:
+            self.qi = qi
+
+    def forward(self, x):
+        if hasattr(self, 'qi'):
+            self.qi.update(x)
+            x = FakeQuantize.apply(x, self.qi)
+
+        x = F.max_pool2d(x, self.kernel_size, self.stride, self.padding)
+
+        return x
+
+    def quantize_inference(self, x):
+        return F.max_pool2d(x, self.kernel_size, self.stride, self.padding)
+

ykl/AlexNet/param_flops.txt

+AlexNet(
+  3.87 M, 100.000% Params, 70.08 MMac, 100.000% MACs, 
+  (conv1): Conv2d(896, 0.023% Params, 917.5 KMac, 1.309% MACs, 3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+  (relu1): ReLU(0, 0.000% Params, 32.77 KMac, 0.047% MACs, inplace=True)
+  (pool1): MaxPool2d(0, 0.000% Params, 32.77 KMac, 0.047% MACs, kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
+  (conv2): Conv2d(18.5 k, 0.478% Params, 4.73 MMac, 6.756% MACs, 32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+  (relu2): ReLU(0, 0.000% Params, 16.38 KMac, 0.023% MACs, inplace=True)
+  (pool2): MaxPool2d(0, 0.000% Params, 16.38 KMac, 0.023% MACs, kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
+  (conv3): Conv2d(73.86 k, 1.909% Params, 4.73 MMac, 6.745% MACs, 64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+  (relu3): ReLU(0, 0.000% Params, 8.19 KMac, 0.012% MACs, inplace=True)
+  (conv4): Conv2d(295.17 k, 7.630% Params, 18.89 MMac, 26.955% MACs, 128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+  (relu4): ReLU(0, 0.000% Params, 16.38 KMac, 0.023% MACs, inplace=True)
+  (conv5): Conv2d(590.08 k, 15.252% Params, 37.77 MMac, 53.887% MACs, 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+  (relu5): ReLU(0, 0.000% Params, 16.38 KMac, 0.023% MACs, inplace=True)
+  (pool5): MaxPool2d(0, 0.000% Params, 16.38 KMac, 0.023% MACs, kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)
+  (drop1): Dropout(0, 0.000% Params, 0.0 Mac, 0.000% MACs, p=0.5, inplace=False)
+  (fc1): Linear(2.36 M, 61.010% Params, 2.36 MMac, 3.368% MACs, in_features=2304, out_features=1024, bias=True)
+  (relu6): ReLU(0, 0.000% Params, 1.02 KMac, 0.001% MACs, inplace=True)
+  (drop2): Dropout(0, 0.000% Params, 0.0 Mac, 0.000% MACs, p=0.5, inplace=False)
+  (fc2): Linear(524.8 k, 13.565% Params, 524.8 KMac, 0.749% MACs, in_features=1024, out_features=512, bias=True)
+  (relu7): ReLU(0, 0.000% Params, 512.0 Mac, 0.001% MACs, inplace=True)
+  (fc3): Linear(5.13 k, 0.133% Params, 5.13 KMac, 0.007% MACs, in_features=512, out_features=10, bias=True)
+)

ykl/AlexNet/ptq.py

+from torch.serialization import load
+from model import *
+from extract_ratio import *
+from utils import *
+
+import gol
+import openpyxl
+import sys
+import argparse
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.optim as optim
+from torchvision import datasets, transforms
+from torchvision.transforms.functional import InterpolationMode
+import os
+import os.path as osp
+from torch.utils.tensorboard import SummaryWriter
+
+
+def direct_quantize(model, test_loader,device):
+    for i, (data, target) in enumerate(test_loader, 1):
+        data = data.to(device)
+        output = model.quantize_forward(data).cpu()
+        if i % 500 == 0:
+            break
+    print('direct quantization finish')
+
+
+def full_inference(model, test_loader, device):
+    correct = 0
+    for i, (data, target) in enumerate(test_loader, 1):
+        data = data.to(device)
+        output = model(data).cpu()
+        pred = output.argmax(dim=1, keepdim=True)
+        # print(pred)
+        correct += pred.eq(target.view_as(pred)).sum().item()
+    print('\nTest set: Full Model Accuracy: {:.2f}%'.format(100. * correct / len(test_loader.dataset)))
+    return 100. * correct / len(test_loader.dataset)
+
+
+def quantize_inference(model, test_loader, device):
+    correct = 0
+    for i, (data, target) in enumerate(test_loader, 1):
+        data = data.to(device)
+        output = model.quantize_inference(data).cpu()
+        pred = output.argmax(dim=1, keepdim=True)
+        correct += pred.eq(target.view_as(pred)).sum().item()
+    print('Test set: Quant Model Accuracy: {:.2f}%'.format(100. * correct / len(test_loader.dataset)))
+    return 100. * correct / len(test_loader.dataset)
+
+
+def js_div(p_output, q_output, get_softmax=True):
+    """
+    Function that measures JS divergence between target and output logits:
+    """
+    KLDivLoss = nn.KLDivLoss(reduction='sum')
+    if get_softmax:
+        p_output = F.softmax(p_output)
+        q_output = F.softmax(q_output)
+    log_mean_output = ((p_output + q_output)/2).log()
+    return (KLDivLoss(log_mean_output, p_output) + KLDivLoss(log_mean_output, q_output))/2
+
+
+if __name__ == "__main__":
+    batch_size = 64
+
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+    print(device)
+    train_loader = torch.utils.data.DataLoader(
+        datasets.CIFAR10('data', train=True, download=True,
+                         transform=transforms.Compose([
+                             transforms.Resize((32, 32), interpolation=InterpolationMode.BICUBIC),
+                             transforms.RandomHorizontalFlip(),
+                             transforms.ToTensor(),
+                             transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
+                         ])),
+        batch_size=batch_size, shuffle=True, num_workers=1, pin_memory=True
+    )
+
+    test_loader = torch.utils.data.DataLoader(
+        datasets.CIFAR10('data', train=False, transform=transforms.Compose([
+            transforms.Resize((32, 32), interpolation=InterpolationMode.BICUBIC),
+            transforms.ToTensor(),
+            transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
+        ])),
+        batch_size=batch_size, shuffle=True, num_workers=1, pin_memory=True
+    )
+
+    model = AlexNet()
+    writer = SummaryWriter(log_dir='./log')
+    full_file = 'ckpt/cifar10_AlexNet.pt'
+    model.load_state_dict(torch.load(full_file))
+    model.to(device)
+
+    load_ptq = True
+    ptq_file_prefix = 'ckpt/cifar10_AlexNet_ptq_'
+
+    model.eval()
+    full_acc = full_inference(model, test_loader, device)
+    full_params = []
+
+    layer, par_ratio, flop_ratio = extract_ratio()
+
+    for name, param in model.named_parameters():
+        param_norm = F.normalize(param.data.cpu(),p=2,dim=-1)
+        full_params.append(param_norm)
+        writer.add_histogram(tag='Full_' + name + '_data', values=param.data)
+
+    gol._init()
+    quant_type_list = ['INT','POT','FLOAT']
+    title_list = []
+    js_flops_list = []
+    js_param_list = []
+    ptq_acc_list = []
+    acc_loss_list = []
+
+    for quant_type in quant_type_list:
+        num_bit_list, e_bit_list = bit_list(quant_type)
+
+        # 对一个量化类别，只需设置一次bias量化表
+        # int由于位宽大，使用量化表开销过大，直接_round即可
+        if quant_type != 'INT':
+            bias_list = build_bias_list(quant_type)
+            gol.set_value(bias_list, is_bias=True)
+
+        for num_bits in num_bit_list:
+            for e_bits in e_bit_list:
+                model_ptq = AlexNet()
+                if quant_type == 'FLOAT':
+                    title = '%s_%d_E%d' % (quant_type, num_bits, e_bits)
+                else:
+                    title = '%s_%d' % (quant_type, num_bits)
+                print('\nPTQ: '+title)
+                title_list.append(title)
+
+                # 设置量化表
+                if quant_type != 'INT':
+                    plist = build_list(quant_type, num_bits, e_bits)
+                    gol.set_value(plist)
+
+                # 判断是否需要载入
+                if load_ptq is True and osp.exists(ptq_file_prefix + title + '.pt'):
+                    model_ptq.quantize(quant_type,num_bits,e_bits)
+                    model_ptq.load_state_dict(torch.load(ptq_file_prefix + title + '.pt'))
+                    model_ptq.to(device)
+                    print('Successfully load ptq model: ' + title)
+                else:
+                    model_ptq.load_state_dict(torch.load(full_file))
+                    model_ptq.to(device)
+                    model_ptq.quantize(quant_type,num_bits,e_bits)
+                    model_ptq.eval()
+                    direct_quantize(model_ptq, train_loader, device)
+                    torch.save(model_ptq.state_dict(), ptq_file_prefix + title + '.pt')
+
+                model_ptq.freeze()
+                ptq_acc = quantize_inference(model_ptq, test_loader, device)
+                ptq_acc_list.append(ptq_acc)
+                acc_loss = (full_acc - ptq_acc) / full_acc
+                acc_loss_list.append(acc_loss)
+                idx = -1
+
+                # 获取计算量/参数量下的js-div
+                js_flops = 0.
+                js_param = 0.
+                for name, param in model_ptq.named_parameters():
+                    if '.' not in name:
+                        continue
+                    idx = idx + 1
+                    prefix = name.split('.')[0]
+                    if prefix in layer:
+                        layer_idx = layer.index(prefix)
+                        ptq_param = param.data.cpu()
+                        # 取L2范数
+                        ptq_norm = F.normalize(ptq_param,p=2,dim=-1)
+                        writer.add_histogram(tag=title +':'+ name + '_data', values=ptq_param)
+                        js = js_div(ptq_norm,full_params[idx])
+                        js = js.item()
+                        if js < 0.:
+                            js = 0.
+                        js_flops = js_flops + js * flop_ratio[layer_idx]
+                        js_param = js_param + js * flop_ratio[layer_idx]
+                js_flops_list.append(js_flops)
+                js_param_list.append(js_param)
+
+                print(title + ': js_flops: %f js_param: %f acc_loss: %f' % (js_flops, js_param, acc_loss))
+    
+    # 写入xlsx
+    workbook = openpyxl.Workbook()
+    worksheet = workbook.active
+    worksheet.cell(row=1,column=1,value='FP32-acc')
+    worksheet.cell(row=1,column=2,value=full_acc)
+    worksheet.cell(row=3,column=1,value='title')
+    worksheet.cell(row=3,column=2,value='js_flops')
+    worksheet.cell(row=3,column=3,value='js_param')
+    worksheet.cell(row=3,column=4,value='ptq_acc')
+    worksheet.cell(row=3,column=5,value='acc_loss')
+    for i in range(len(title_list)):
+        worksheet.cell(row=i+4, column=1, value=title_list[i])
+        worksheet.cell(row=i+4, column=2, value=js_flops_list[i])
+        worksheet.cell(row=i+4, column=3, value=js_param_list[i])
+        worksheet.cell(row=i+4, column=4, value=ptq_acc_list[i])
+        worksheet.cell(row=i+4, column=5, value=acc_loss_list[i])
+    
+    workbook.save('ptq_result.xlsx')
+    writer.close()
+
+    ft = open('ptq_result.txt','w')
+
+    print('title_list:',file=ft)
+    print(" ".join(title_list),file=ft)
+    print('js_flops_list:',file=ft)
+    print(" ".join(str(i) for i in js_flops_list), file=ft)
+    print('js_param_list:',file=ft)
+    print(" ".join(str(i) for i in js_param_list), file=ft)
+    print('ptq_acc_list:',file=ft)
+    print(" ".join(str(i) for i in ptq_acc_list), file=ft)
+    print('acc_loss_list:',file=ft)
+    print(" ".join(str(i) for i in acc_loss_list), file=ft)
+
+    ft.close()
+

ykl/AlexNet/ptq_result.txt

+title_list:
+INT_2 INT_3 INT_4 INT_5 INT_6 INT_7 INT_8 INT_9 INT_10 INT_11 INT_12 INT_13 INT_14 INT_15 INT_16 POT_2 POT_3 POT_4 POT_5 POT_6 POT_7 POT_8 FLOAT_8_E1 FLOAT_8_E2 FLOAT_8_E3 FLOAT_8_E4 FLOAT_8_E5 FLOAT_8_E6
+js_flops_list:
+7507.750226317713 2739.698390971301 602.5613310246055 140.92197221503724 34.51721888016634 8.518508718865842 2.1353732883428638 0.5319411628570782 0.1316271020831477 0.03249564657892055 0.008037284252895557 0.0020460099353784723 0.00041867764927864105 0.0001321614950419231 5.841430176387608e-06 7507.667348902921 1654.3775934528933 136.7401730898288 134.5782970456457 134.57841422062364 134.5783939274636 134.5782945727605 33.31638902152266 32.12034308540418 0.6541880874259414 2.442034364817909 9.688117360231624 37.70544899186622
+js_param_list:
+7507.750226317713 2739.698390971301 602.5613310246055 140.92197221503724 34.51721888016634 8.518508718865842 2.1353732883428638 0.5319411628570782 0.1316271020831477 0.03249564657892055 0.008037284252895557 0.0020460099353784723 0.00041867764927864105 0.0001321614950419231 5.841430176387608e-06 7507.667348902921 1654.3775934528933 136.7401730898288 134.5782970456457 134.57841422062364 134.5783939274636 134.5782945727605 33.31638902152266 32.12034308540418 0.6541880874259414 2.442034364817909 9.688117360231624 37.70544899186622
+ptq_acc_list:
+10.0 10.16 51.21 77.39 83.03 84.73 84.84 85.01 85.08 85.07 85.06 85.08 85.08 85.08 85.08 10.0 14.32 72.49 72.65 72.95 72.08 72.23 82.73 83.3 85.01 84.77 59.86 51.87
+acc_loss_list:
+0.8824635637047484 0.8805829807240245 0.3980959097320169 0.0903855195110484 0.02409496944052653 0.004113775270333736 0.0028208744710859768 0.0008227550540666805 0.0 0.00011753643629531167 0.0002350728725904563 0.0 0.0 0.0 0.0 0.8824635637047484 0.8316878232251997 0.14797837329572172 0.14609779031499756 0.14257169722614005 0.152797367183827 0.15103432063939815 0.027621062529384042 0.020921485660554785 0.0008227550540666805 0.0036436295251528242 0.29642689233662434 0.39033850493653033

ykl/AlexNet/qat.py

+from model import *
+from utils import *
+import gol
+import sys
+import argparse
+import torch
+import torch.nn as nn
+import torch.optim as optim
+from torchvision import datasets, transforms
+import os
+import os.path as osp
+
+
+def quantize_aware_training(model, device, train_loader, optimizer, epoch):
+    lossLayer = torch.nn.CrossEntropyLoss()
+    for batch_idx, (data, target) in enumerate(train_loader, 1):
+        data, target = data.to(device), target.to(device)
+        optimizer.zero_grad()
+        output = model.quantize_forward(data)
+        loss = lossLayer(output, target)
+        loss.backward()
+        optimizer.step()
+
+        if batch_idx % 50 == 0:
+            print('Quantize Aware Training Epoch: {} [{}/{}]\tLoss: {:.6f}'.format(
+                epoch, batch_idx * len(data), len(train_loader.dataset), loss.item()
+            ))
+
+
+def full_inference(model, test_loader):
+    correct = 0
+    for i, (data, target) in enumerate(test_loader, 1):
+        data, target = data.to(device), target.to(device)
+        output = model(data)
+        pred = output.argmax(dim=1, keepdim=True)
+        correct += pred.eq(target.view_as(pred)).sum().item()
+    print('\nTest set: Full Model Accuracy: {:.2f}%\n'.format(100. * correct / len(test_loader.dataset)))
+
+
+def quantize_inference(model, test_loader):
+    correct = 0
+    for i, (data, target) in enumerate(test_loader, 1):
+        data, target = data.to(device), target.to(device)
+        output = model.quantize_inference(data)
+        pred = output.argmax(dim=1, keepdim=True)
+        correct += pred.eq(target.view_as(pred)).sum().item()
+    print('\nTest set: Quant Model Accuracy: {:.2f}%\n'.format(100. * correct / len(test_loader.dataset)))
+
+
+if __name__ == "__main__":
+    # arg1 = int(sys.argv[1])  # epoch
+    # arg2 = int(sys.argv[2])  # bits of quantization
+
+    batch_size = 32
+    seed = 1
+    epochs1 = 3
+    epochs2 = 30 # 16~30
+    lr1 = 0.01
+    lr2 = 0.001
+    momentum = 0.5
+    using_bn = False
+
+    torch.manual_seed(seed)
+
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+
+    train_loader = torch.utils.data.DataLoader(
+        datasets.CIFAR10('data', train=True, download=True,
+                       transform=transforms.Compose([
+                            transforms.ToTensor(),
+                            transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
+                       ])),
+        batch_size=batch_size, shuffle=True, num_workers=1, pin_memory=False
+    )
+
+    test_loader = torch.utils.data.DataLoader(
+        datasets.CIFAR10('data', train=False, transform=transforms.Compose([
+            transforms.ToTensor(),
+            transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
+        ])),
+        batch_size=batch_size, shuffle=True, num_workers=1, pin_memory=False
+    )
+
+    # if using_bn:
+    #     model = NetBN()
+    #     model.load_state_dict(torch.load('ckpt/mnist_cnnbn.pt', map_location='cpu'))
+    #     save_file = "ckpt/mnist_cnnbn_qat.pt"
+    # else:
+    #     model = Net()
+    #     model.load_state_dict(torch.load('ckpt/mnist_cnn.pt', map_location='cpu'))
+    #     save_file = "ckpt/mnist_cnn_qat.pt"
+
+    model = AlexNet()
+#     model.load_state_dict(torch.load('ckpt/cifar10_AlexNet_t5.pt', map_location='cpu'))
+    model.load_state_dict(torch.load('ckpt/cifar10_AlexNet_t4.pt'))
+    save_file = "ckpt/cifar10_AlexNet_qat_e4.pt"
+    load_quant_model_file = None
+    # load_quant_model_file = "ckpt/cifar10_AlexNet_qat_ratio_4.pt"
+    model.to(device)
+
+    # 原来是所有参数，包括原有conv的参数
+    # optimizer = optim.SGD(model.parameters(), lr=lr, momentum=momentum)
+    # 改动后只训练每层的scale，需要在量化后才能指定
+    
+    model.eval()
+    
+    full_inference(model, test_loader)
+    # for param_tensor, param_value in model.state_dict().items():
+    #     print(param_tensor, "\t", param_value)
+    num_bits = 8
+    e_bits = 4
+    gol._init()
+    plist = build_list(num_bits=16,e_bits=5)
+    gol.set_value(plist,is_bias=True)
+
+    plist = build_list(num_bits,e_bits)
+    gol.set_value(plist)
+    plist = build_list(num_bits,e_bits)
+    # print(plist)
+    gol.set_value(plist)
+
+    model.quantize(num_bits,e_bits)
+    print('Quantization bit: %d' % num_bits)
+
+    params,params_name = model.get_quant_scales()
+    optimizer1 = optim.SGD(params, lr=lr1, momentum=momentum)
+    optimizer2 = optim.SGD(params, lr=lr2, momentum=momentum)
+    # print('--debug--')
+    # for name in params_name:
+    #     print(name)
+    # input()
+
+    if load_quant_model_file is not None:
+        model.load_state_dict(torch.load(load_quant_model_file))
+        print("Successfully load quantized model %s" % load_quant_model_file)
+    else:
+        model.train()
+
+        for epoch in range(1, epochs1 + 1):
+            quantize_aware_training(model, device, train_loader, optimizer1, epoch)
+        # for epoch in range(epochs1 + 1, epochs2 + 1):
+        #     quantize_aware_training(model, device, train_loader, optimizer2, epoch)
+
+        model.eval()
+        torch.save(model.state_dict(), save_file)
+
+    model.freeze()
+    # for name, param in model.named_parameters():
+    #     print(name)
+    #     print(param.data)
+    #     print('----------')
+
+    # for param_tensor, param_value in model.state_dict().items():
+    #     print(param_tensor, "\t", param_value)
+    quantize_inference(model, test_loader)
+
+    
+
+
+
+    

ykl/AlexNet/train.py

+from model import *
+
+import torch
+import torch.nn as nn
+import torch.optim as optim
+from torchvision import datasets, transforms
+from torchvision.transforms.functional import InterpolationMode
+import os
+import os.path as osp
+
+
+def train(model, device, train_loader, optimizer, epoch):
+    model.train()
+    lossLayer = torch.nn.CrossEntropyLoss()
+    for batch_idx, (data, target) in enumerate(train_loader):
+        data, target = data.to(device), target.to(device)
+        optimizer.zero_grad()
+        output = model(data)
+        loss = lossLayer(output, target)
+        loss.backward()
+        optimizer.step()
+
+        if batch_idx % 50 == 0:
+            print('Train Epoch: {} [{}/{}]\tLoss: {:.6f}'.format(
+                epoch, batch_idx * len(data), len(train_loader.dataset), loss.item()
+            ))
+
+def test(model, device, test_loader):
+    model.eval()
+    test_loss = 0
+    correct = 0
+    lossLayer = torch.nn.CrossEntropyLoss(reduction='sum')
+    for data, target in test_loader:
+        data, target = data.to(device), target.to(device)
+        output = model(data)
+        test_loss += lossLayer(output, target).item()
+        pred = output.argmax(dim=1, keepdim=True)
+        correct += pred.eq(target.view_as(pred)).sum().item()
+    
+    test_loss /= len(test_loader.dataset)
+
+    print('\nTest set: Average loss: {:.4f}, Accuracy: {:.2f}%\n'.format(
+        test_loss, 100. * correct / len(test_loader.dataset)
+    ))
+
+
+if __name__ == "__main__":
+    batch_size = 32
+    test_batch_size = 32
+    seed = 1
+    epochs1 = 15
+    epochs2 = epochs1+10
+    epochs3 = epochs2+10
+    lr1 = 0.01
+    lr2 = 0.001
+    lr3 = 0.0001
+    momentum = 0.5
+    save_model = True
+
+    torch.manual_seed(seed)
+
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+
+    train_loader = torch.utils.data.DataLoader(
+        datasets.CIFAR10('data', train=True, download=True,
+                       transform=transforms.Compose([
+                           transforms.Resize((32, 32), interpolation=InterpolationMode.BICUBIC),
+                           transforms.RandomHorizontalFlip(),
+                           transforms.ToTensor(),
+                           transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
+                       ])),
+        batch_size=batch_size, shuffle=True, num_workers=1, pin_memory=True
+    )
+
+    test_loader = torch.utils.data.DataLoader(
+        datasets.CIFAR10('data', train=False, transform=transforms.Compose([
+            transforms.Resize((32, 32), interpolation=InterpolationMode.BICUBIC),
+            transforms.ToTensor(),
+            transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
+        ])),
+        batch_size=test_batch_size, shuffle=True, num_workers=1, pin_memory=True
+    )
+
+    model = AlexNet().to(device)
+
+    optimizer1 = optim.SGD(model.parameters(), lr=lr1, momentum=momentum)
+    optimizer2 = optim.SGD(model.parameters(), lr=lr2, momentum=momentum)
+    optimizer3 = optim.SGD(model.parameters(), lr=lr3, momentum=momentum)
+    
+    for epoch in range(1, epochs1 + 1):
+        train(model, device, train_loader, optimizer1, epoch)
+        test(model, device, test_loader)
+    
+    for epoch in range(epochs1 + 1, epochs2 + 1):
+        train(model, device, train_loader, optimizer2, epoch)
+        test(model, device, test_loader)
+        
+    for epoch in range(epochs2 + 1, epochs3 + 1):
+        train(model, device, train_loader, optimizer3, epoch)
+        test(model, device, test_loader)
+    
+    if save_model:
+        if not osp.exists('ckpt'):
+            os.makedirs('ckpt')
+        torch.save(model.state_dict(), 'ckpt/cifar10_AlexNet.pt')
\ No newline at end of file

ykl/AlexNet/utils.py

+import torch
+
+def bit_list(quant_type):
+    if quant_type == 'INT':
+        num_bit_list = list(range(2,17))
+        e_bit_list = [0]
+    elif quant_type == 'POT':
+        num_bit_list = list(range(2,9))
+        e_bit_list = [0]
+    else:
+        num_bit_list = [8]
+        e_bit_list = list(range(1,7))
+    
+    return num_bit_list, e_bit_list     
+
+def build_bias_list(quant_type):
+    if quant_type == 'POT':
+        return build_pot_list(8)
+    else:
+        return build_float_list(16,7)
+    
+def build_list(quant_type, num_bits, e_bits):
+    if quant_type == 'POT':
+        return build_pot_list(num_bits)
+    else:
+        return build_float_list(num_bits,e_bits)
+
+def build_pot_list(num_bits):
+    plist = [0.]
+    for i in range(-2 ** (num_bits-1) + 2, 1): 
+        # i最高到0，即pot量化最大值为1
+        plist.append(2. ** i)
+        plist.append(-2. ** i)
+    plist = torch.Tensor(list(set(plist)))
+    # plist = plist.mul(1.0 / torch.max(plist))
+    return plist
+
+def build_float_list(num_bits,e_bits):
+    m_bits = num_bits - 1 - e_bits
+    plist = [0.]
+    # 相邻尾数的差值
+    dist_m = 2 ** (-m_bits)
+    e = -2 ** (e_bits - 1) + 1
+    for m in range(1, 2 ** m_bits):
+        frac = m * dist_m   # 尾数部分
+        expo = 2 ** e       # 指数部分
+        flt = frac * expo
+        plist.append(flt)
+        plist.append(-flt)
+
+    for e in range(-2 ** (e_bits - 1) + 2, 2 ** (e_bits - 1) + 1):
+        expo = 2 ** e
+        for m in range(0, 2 ** m_bits):
+            frac = 1. + m * dist_m
+            flt = frac * expo
+            plist.append(flt)
+            plist.append(-flt)
+    plist = torch.Tensor(list(set(plist)))
+    return plist
\ No newline at end of file