repvgg.py

import torch.nn as nn
import numpy as np
import torch


def deploy(self, mode=False,show_error = False):
    self.deploying = mode
    for module in self.children():
        if hasattr(module, 'deploying'):
            module.deploy(mode,show_error)
nn.Sequential.deploying = False
nn.Sequential.show_error = False
nn.Sequential.deploy = deploy

def conv_bn(in_channels, out_channels, kernel_size, stride, padding, groups=1):
    result = nn.Sequential()
    result.add_module('conv', nn.Conv2d(in_channels=in_channels, out_channels=out_channels,
                                                  kernel_size=kernel_size, stride=stride, padding=padding, groups=groups, bias=False))
    result.add_module('bn', nn.BatchNorm2d(num_features=out_channels))
    return result

class RepVGGBlock(nn.Module):

    def __init__(self, in_channels, out_channels, kernel_size,
                 stride=1, padding=0, groups=1):
        super(RepVGGBlock, self).__init__()
        self.deploying = False
        self.show_error  = False

        self.groups = groups
        self.in_channels = in_channels

        assert kernel_size == 3
        assert padding == 1

        padding_11 = padding - kernel_size // 2

        self.nonlinearity = nn.ReLU()

        self.in_channels = in_channels
        self.in_channels = in_channels
        self.kernel_size = kernel_size
        self.stride = stride
        self.padding = padding
        self.groups = groups

        self.register_buffer('fused_weight', None)
        self.register_buffer('fused_bias', None)

        self.rbr_identity = nn.BatchNorm2d(num_features=in_channels) if out_channels == in_channels and stride == 1 else None
        self.rbr_dense = conv_bn(in_channels=in_channels, out_channels=out_channels, kernel_size=kernel_size, stride=stride, padding=padding, groups=groups)
        self.rbr_1x1 = conv_bn(in_channels=in_channels, out_channels=out_channels, kernel_size=1, stride=stride, padding=padding_11, groups=groups)
        print('RepVGG Block, identity = ', self.rbr_identity)


    def forward(self, inputs):
        if self.deploying :
            if self.fused_weight is None or self.fused_bias is None:
                self.fused_weight,self.fused_bias = self.get_equivalent_kernel_bias()

            fused_out =  self.nonlinearity(torch.nn.functional.conv2d(
                        inputs,self.fused_weight,self.fused_bias,self.stride,self.padding,1,self.groups))

            if not self.show_error:
                return fused_out

        if self.rbr_identity is None:
            id_out = 0
        else:
            id_out = self.rbr_identity(inputs)
        out = self.nonlinearity(self.rbr_dense(inputs) + self.rbr_1x1(inputs) + id_out)

        if self.deploying and self.show_error:
            print(torch.max(torch.abs(fused_out - out)).item())
            return fused_out

        return out

#   This func derives the equivalent kernel and bias in a DIFFERENTIABLE way.
#   You can get the equivalent kernel and bias at any time and do whatever you want,
    #   for example, apply some penalties or constraints during training, just like you do to the other models.
#   May be useful for quantization or pruning.
    def get_equivalent_kernel_bias(self):
        kernel3x3, bias3x3 = self._fuse_bn_tensor(self.rbr_dense)
        kernel1x1, bias1x1 = self._fuse_bn_tensor(self.rbr_1x1)
        kernelid, biasid = self._fuse_bn_tensor(self.rbr_identity)
        return kernel3x3 + self._pad_1x1_to_3x3_tensor(kernel1x1) + kernelid, bias3x3 + bias1x1 + biasid

    def _pad_1x1_to_3x3_tensor(self, kernel1x1):
        if kernel1x1 is None:
            return 0
        else:
            return torch.nn.functional.pad(kernel1x1, [1,1,1,1])

    def _fuse_bn_tensor(self, branch):
        if branch is None:
            return 0, 0
        if isinstance(branch, nn.Sequential):
            kernel = branch.conv.weight
            running_mean = branch.bn.running_mean
            running_var = branch.bn.running_var
            gamma = branch.bn.weight
            beta = branch.bn.bias
            eps = branch.bn.eps
        else:
            assert isinstance(branch, nn.BatchNorm2d)
            if not hasattr(self, 'id_tensor'):
                input_dim = self.in_channels // self.groups
                kernel_value = np.zeros((self.in_channels, input_dim, 3, 3), dtype=np.float32)
                for i in range(self.in_channels):
                    kernel_value[i, i % input_dim, 1, 1] = 1
                self.id_tensor = torch.from_numpy(kernel_value).to(branch.weight.device)
            kernel = self.id_tensor
            running_mean = branch.running_mean
            running_var = branch.running_var
            gamma = branch.weight
            beta = branch.bias
            eps = branch.eps
        std = (running_var + eps).sqrt()
        t = (gamma / std).reshape(-1, 1, 1, 1)
        return kernel * t, beta - running_mean * gamma / std

    def deploy(self,mode = False,show_error = False):
        self.deploying = mode
        self.show_error = show_error


class RepVGG(nn.Module):

    def __init__(self, num_blocks, num_classes=1000, width_multiplier=None, override_groups_map=None):
        super(RepVGG, self).__init__()

        assert len(width_multiplier) == 4

        self.deploying = deploy
        self.show_error = False

        self.override_groups_map = override_groups_map or dict()

        assert 0 not in self.override_groups_map

        self.in_planes = min(64, int(64 * width_multiplier[0]))

        self.stage0 = RepVGGBlock(in_channels=3, out_channels=self.in_planes, kernel_size=3, stride=2, padding=1)
        self.cur_layer_idx = 1
        self.stage1 = self._make_stage(int(64 * width_multiplier[0]), num_blocks[0], stride=2)
        self.stage2 = self._make_stage(int(128 * width_multiplier[1]), num_blocks[1], stride=2)
        self.stage3 = self._make_stage(int(256 * width_multiplier[2]), num_blocks[2], stride=2)
        self.stage4 = self._make_stage(int(512 * width_multiplier[3]), num_blocks[3], stride=2)
        self.gap = nn.AdaptiveAvgPool2d(output_size=1)
        self.linear = nn.Linear(int(512 * width_multiplier[3]), num_classes)


    def _make_stage(self, planes, num_blocks, stride):
        strides = [stride] + [1]*(num_blocks-1)
        blocks = []
        for stride in strides:
            cur_groups = self.override_groups_map.get(self.cur_layer_idx, 1)
            blocks.append(RepVGGBlock(in_channels=self.in_planes, out_channels=planes, kernel_size=3,
                                      stride=stride, padding=1, groups=cur_groups))
            self.in_planes = planes
            self.cur_layer_idx += 1
        return nn.Sequential(*blocks)

    def deploy(self,mode = False,show_error = False):
        self.deploying = mode
        for module in self.children():
            if hasattr(module,'deploying'):
                module.deploy(mode,show_error)


    def forward(self, x):
        out = self.stage0(x)
        out = self.stage1(out)
        out = self.stage2(out)
        out = self.stage3(out)
        out = self.stage4(out)
        out = self.gap(out)
        out = out.view(out.size(0), -1)
        out = self.linear(out)
        return out




optional_groupwise_layers = [2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26]
g2_map = {l: 2 for l in optional_groupwise_layers}
g4_map = {l: 4 for l in optional_groupwise_layers}

def create_RepVGG_A0():
    return RepVGG(num_blocks=[2, 4, 14, 1], num_classes=1000,
                  width_multiplier=[0.75, 0.75, 0.75, 2.5], override_groups_map=None)

def create_RepVGG_A1():
    return RepVGG(num_blocks=[2, 4, 14, 1], num_classes=1000,
                  width_multiplier=[1, 1, 1, 2.5], override_groups_map=None)

def create_RepVGG_A2():
    return RepVGG(num_blocks=[2, 4, 14, 1], num_classes=1000,
                  width_multiplier=[1.5, 1.5, 1.5, 2.75], override_groups_map=None)

def create_RepVGG_B0():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[1, 1, 1, 2.5], override_groups_map=None)

def create_RepVGG_B1():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[2, 2, 2, 4], override_groups_map=None)

def create_RepVGG_B1g2():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[2, 2, 2, 4], override_groups_map=g2_map)

def create_RepVGG_B1g4():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[2, 2, 2, 4], override_groups_map=g4_map)


def create_RepVGG_B2():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=None)

def create_RepVGG_B2g2():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=g2_map)

def create_RepVGG_B2g4():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=g4_map)


def create_RepVGG_B3():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[3, 3, 3, 5], override_groups_map=None)

def create_RepVGG_B3g2():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[3, 3, 3, 5], override_groups_map=g2_map)

def create_RepVGG_B3g4():
    return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
                  width_multiplier=[3, 3, 3, 5], override_groups_map=g4_map)


func_dict = {
'RepVGG-A0': create_RepVGG_A0,
'RepVGG-A1': create_RepVGG_A1,
'RepVGG-A2': create_RepVGG_A2,
'RepVGG-B0': create_RepVGG_B0,
'RepVGG-B1': create_RepVGG_B1,
'RepVGG-B1g2': create_RepVGG_B1g2,
'RepVGG-B1g4': create_RepVGG_B1g4,
'RepVGG-B2': create_RepVGG_B2,
'RepVGG-B2g2': create_RepVGG_B2g2,
'RepVGG-B2g4': create_RepVGG_B2g4,
'RepVGG-B3': create_RepVGG_B3,
'RepVGG-B3g2': create_RepVGG_B3g2,
'RepVGG-B3g4': create_RepVGG_B3g4,
}
def get_RepVGG_func_by_name(name):
    return func_dict[name]


if __name__ == '__main__':
    model = get_RepVGG_func_by_name("RepVGG-A0")()
    model.eval()

    test_in = torch.rand([1,3,224,224])

    y = model(test_in)
    model.deploy(mode=True,show_error=False)

    fused_y = model(test_in)

    print("final error :", torch.max(torch.abs(fused_y - y )).item())

    torch.onnx.export(model,test_in,"model.onnx")