ackbone：EfficientViT
论文地址：https://openaccess.thecvf.com/content/CVPR2023/papers/Liu_EfficientViT_Memory_Efficient_Vision_Transformer_With_Cascaded_Group_Attention_CVPR_2023_paper.pdf
代码仓库：https://github.com/microsoft/Cream/tree/main/EfficientViT
摘要：视觉transformer由于其高模型能力而取得了巨大的成功。然而，它们卓越的性能伴随着沉重的计算成本，这使得它们不适合实时应用。在这篇论文中，本文提出了一个高速视觉transformer家族，名为EfficientViT。本文发现现有的transformer模型的速度通常受到内存低效操作的限制，特别是在MHSA中的张量重塑和单元函数。因此，本文设计了一种具有三明治布局的新构建块，即在高效FFN层之间使用单个内存绑定的MHSA，从而提高了内存效率，同时增强了信道通信。此外，本文发现注意图在头部之间具有很高的相似性，从而导致计算冗余。为了解决这个问题，本文提出了一个级联的群体注意模块，以不同的完整特征分割来馈送注意头，不仅节省了计算成本，而且提高了注意多样性。综合实验表明，高效vit优于现有的高效模型，在速度和精度之间取得了良好的平衡。例如，本文的EfficientViT-M5在准确率上比MobileNetV3-Large高出1.9%，而在Nvidia V100 GPU和Intel Xeon CPU上的吞吐量分别高出40.4%和45.2%。与最近的高效型号MobileViT-XXS相比，efficientvitt - m2的精度提高了1.8%，同时在GPU/CPU上运行速度提高了5.8 ×/3.7 ×，转换为ONNX格式时速度提高了7.4×。
本文通过分析DeiT和Swin两个Transformer架构得出如下结论：
-
适当降低MHSA层利用率可以在提高模型性能的同时提高访存效率
-
在不同的头部使用不同的通道划分特征，而不是像MHSA那样对所有头部使用相同的全特征，可以有效地减少注意力计算冗余
-
典型的通道配置，即在每个阶段之后将通道数加倍或对所有块使用等效通道，可能在最后几个块中产生大量冗余
-
在维度相同的情况下，Q、K的冗余度比V大得多
-
a new building block with a sandwich layout（减少self-attention的次数）：之前是一个block self-attention->fc->self-attention->fc->self-attention->fc->...N次数；现在是一个block fc->self-attention->fc；不仅能够提升内存效率而且能够增强通道间的计算
-
cascaded group attention：让多头串联学习特征：第一个头学习完特征后，第二个头利用第一个头学习到的特征的基础上再去学习(原来的transformer是第二个头跟第一个头同时独立地去学习)，同理第三个头学习时也得利用上第二个头学习的结果再去学习
Efficientvit模型结构如下图所示：
a memory-efficient sandwich layout
Cascaded Group Attention：解决了原来模型中多头重复学习(学习到的特征很多都是相似的)的问题，这里每个头学到的特征都不同，而且越往下面的头学到的特征越丰富。
-
Q是主动查询的行为，特征比K更加丰富，所以额外做了个Token Interation
-
Q进行self-attention之前先通过多次分组卷积再一次学习
Efficientvit变体模型结构如下表所示：
EfficientViT结构代码
# --------------------------------------------------------
# EfficientViT Model Architecture for Downstream Tasks
# Copyright (c) 2022 Microsoft
# Written by: Xinyu Liu
# --------------------------------------------------------
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as checkpoint
import itertools
from timm.models.layers import SqueezeExcite
import numpy as np
import itertools
__all__ = ['EfficientViT_M0', 'EfficientViT_M1', 'EfficientViT_M2', 'EfficientViT_M3', 'EfficientViT_M4', 'EfficientViT_M5']
class Conv2d_BN(torch.nn.Sequential):
def __init__(self, a, b, ks=1, stride=1, pad=0, dilation=1,
groups=1, bn_weight_init=1, resolution=-10000):
super().__init__()
self.add_module('c', torch.nn.Conv2d(
a, b, ks, stride, pad, dilation, groups, bias=False))
self.add_module('bn', torch.nn.BatchNorm2d(b))
torch.nn.init.constant_(self.bn.weight, bn_weight_init)
torch.nn.init.constant_(self.bn.bias, 0)
@torch.no_grad()
def fuse(self):
c, bn = self._modules.values()
w = bn.weight / (bn.running_var + bn.eps)**0.5
w = c.weight * w[:, None, None, None]
b = bn.bias - bn.running_mean * bn.weight / \
(bn.running_var + bn.eps)**0.5
m = torch.nn.Conv2d(w.size(1) * self.c.groups, w.size(
0), w.shape[2:], stride=self.c.stride, padding=self.c.padding, dilation=self.c.dilation, groups=self.c.groups)
m.weight.data.copy_(w)
m.bias.data.copy_(b)
return m
def replace_batchnorm(net):
for child_name, child in net.named_children():
if hasattr(child, 'fuse'):
setattr(net, child_name, child.fuse())
elif isinstance(child, torch.nn.BatchNorm2d):
setattr(net, child_name, torch.nn.Identity())
else:
replace_batchnorm(child)
class PatchMerging(torch.nn.Module):
def __init__(self, dim, out_dim, input_resolution):
super().__init__()
hid_dim = int(dim * 4)
self.conv1 = Conv2d_BN(dim, hid_dim, 1, 1, 0, resolution=input_resolution)
self.act = torch.nn.ReLU()
self.conv2 = Conv2d_BN(hid_dim, hid_dim, 3, 2, 1, groups=hid_dim, resolution=input_resolution)
self.se = SqueezeExcite(hid_dim, .25)
self.conv3 = Conv2d_BN(hid_dim, out_dim, 1, 1, 0, resolution=input_resolution // 2)
def forward(self, x):
x = self.conv3(self.se(self.act(self.conv2(self.act(self.conv1(x))))))
return x
class Residual(torch.nn.Module):
def __init__(self, m, drop=0.):
super().__init__()
self.m = m
self.drop = drop
def forward(self, x):
if self.training and self.drop > 0:
return x + self.m(x) * torch.rand(x.size(0), 1, 1, 1,
device=x.device).ge_(self.drop).div(1 - self.drop).detach()
else:
return x + self.m(x)
class FFN(torch.nn.Module):
def __init__(self, ed, h, resolution):
super().__init__()
self.pw1 = Conv2d_BN(ed, h, resolution=resolution)
self.act = torch.nn.ReLU()
self.pw2 = Conv2d_BN(h, ed, bn_weight_init=0, resolution=resolution)
def forward(self, x):
x = self.pw2(self.act(self.pw1(x)))
return x
class CascadedGroupAttention(torch.nn.Module):
r""" Cascaded Group Attention.
Args:
dim (int): Number of input channels.
key_dim (int): The dimension for query and key.
num_heads (int): Number of attention heads.
attn_ratio (int): Multiplier for the query dim for value dimension.
resolution (int): Input resolution, correspond to the window size.
kernels (List[int]): The kernel size of the dw conv on query.
"""
def __init__(self, dim, key_dim, num_heads=8,
attn_ratio=4,
resolution=14,
kernels=[5, 5, 5, 5],):
super().__init__()
self.num_heads = num_heads
self.scale = key_dim ** -0.5
self.key_dim = key_dim
self.d = int(attn_ratio * key_dim)
self.attn_ratio = attn_ratio
qkvs = []
dws = []
for i in range(num_heads):
qkvs.append(Conv2d_BN(dim // (num_heads), self.key_dim * 2 + self.d, resolution=resolution))
dws.append(Conv2d_BN(self.key_dim, self.key_dim, kernels[i], 1, kernels[i]//2, groups=self.key_dim, resolution=resolution))
self.qkvs = torch.nn.ModuleList(qkvs)
self.dws = torch.nn.ModuleList(dws)
self.proj = torch.nn.Sequential(torch.nn.ReLU(), Conv2d_BN(
self.d * num_heads, dim, bn_weight_init=0, resolution=resolution))
points = list(itertools.product(range(resolution), range(resolution)))
N = len(points)
attention_offsets = {}
idxs = []
for p1 in points:
for p2 in points:
offset = (abs(p1[0] - p2[0]), abs(p1[1] - p2[1]))
if offset not in attention_offsets:
attention_offsets[offset] = len(attention_offsets)
idxs.append(attention_offsets[offset])
self.attention_biases = torch.nn.Parameter(
torch.zeros(num_heads, len(attention_offsets)))
self.register_buffer('attention_bias_idxs',
torch.LongTensor(idxs).view(N, N))
@torch.no_grad()
def train(self, mode=True):
super().train(mode)
if mode and hasattr(self, 'ab'):
del self.ab
else:
self.ab = self.attention_biases[:, self.attention_bias_idxs]
def forward(self, x): # x (B,C,H,W)
B, C, H, W = x.shape
trainingab = self.attention_biases[:, self.attention_bias_idxs]
feats_in = x.chunk(len(self.qkvs), dim=1)
feats_out = []
feat = feats_in[0]
for i, qkv in enumerate(self.qkvs):
if i > 0: # add the previous output to the input
feat = feat + feats_in[i]
feat = qkv(feat)
q, k, v = feat.view(B, -1, H, W).split([self.key_dim, self.key_dim, self.d], dim=1) # B, C/h, H, W
q = self.dws[i](q)
q, k, v = q.flatten(2), k.flatten(2), v.flatten(2) # B, C/h, N
attn = (
(q.transpose(-2, -1) @ k) * self.scale
+
(trainingab[i] if self.training else self.ab[i])
)
attn = attn.softmax(dim=-1) # BNN
feat = (v @ attn.transpose(-2, -1)).view(B, self.d, H, W) # BCHW
feats_out.append(feat)
x = self.proj(torch.cat(feats_out, 1))
return x
class LocalWindowAttention(torch.nn.Module):
r""" Local Window Attention.
Args:
dim (int): Number of input channels.
key_dim (int): The dimension for query and key.
num_heads (int): Number of attention heads.
attn_ratio (int): Multiplier for the query dim for value dimension.
resolution (int): Input resolution.
window_resolution (int): Local window resolution.
kernels (List[int]): The kernel size of the dw conv on query.
"""
def __init__(self, dim, key_dim, num_heads=8,
attn_ratio=4,
resolution=14,
window_resolution=7,
kernels=[5, 5, 5, 5],):
super().__init__()
self.dim = dim
self.num_heads = num_heads
self.resolution = resolution
assert window_resolution > 0, 'window_size must be greater than 0'
self.window_resolution = window_resolution
self.attn = CascadedGroupAttention(dim, key_dim, num_heads,
attn_ratio=attn_ratio,
resolution=window_resolution,
kernels=kernels,)
def forward(self, x):
B, C, H, W = x.shape
if H <= self.window_resolution and W <= self.window_resolution:
x = self.attn(x)
else:
x = x.permute(0, 2, 3, 1)
pad_b = (self.window_resolution - H %
self.window_resolution) % self.window_resolution
pad_r = (self.window_resolution - W %
self.window_resolution) % self.window_resolution
padding = pad_b > 0 or pad_r > 0
if padding:
x = torch.nn.functional.pad(x, (0, 0, 0, pad_r, 0, pad_b))
pH, pW = H + pad_b, W + pad_r
nH = pH // self.window_resolution
nW = pW // self.window_resolution
# window partition, BHWC -> B(nHh)(nWw)C -> BnHnWhwC -> (BnHnW)hwC -> (BnHnW)Chw
x = x.view(B, nH, self.window_resolution, nW, self.window_resolution, C).transpose(2, 3).reshape(
B * nH * nW, self.window_resolution, self.window_resolution, C
).permute(0, 3, 1, 2)
x = self.attn(x)
# window reverse, (BnHnW)Chw -> (BnHnW)hwC -> BnHnWhwC -> B(nHh)(nWw)C -> BHWC
x = x.permute(0, 2, 3, 1).view(B, nH, nW, self.window_resolution, self.window_resolution,
C).transpose(2, 3).reshape(B, pH, pW, C)
if padding:
x = x[:, :H, :W].contiguous()
x = x.permute(0, 3, 1, 2)
return x
class EfficientViTBlock(torch.nn.Module):
""" A basic EfficientViT building block.
Args:
type (str): Type for token mixer. Default: 's' for self-attention.
ed (int): Number of input channels.
kd (int): Dimension for query and key in the token mixer.
nh (int): Number of attention heads.
ar (int): Multiplier for the query dim for value dimension.
resolution (int): Input resolution.
window_resolution (int): Local window resolution.
kernels (List[int]): The kernel size of the dw conv on query.
"""
def __init__(self, type,
ed, kd, nh=8,
ar=4,
resolution=14,
window_resolution=7,
kernels=[5, 5, 5, 5],):
super().__init__()
self.dw0 = Residual(Conv2d_BN(ed, ed, 3, 1, 1, groups=ed, bn_weight_init=0., resolution=resolution))
self.ffn0 = Residual(FFN(ed, int(ed * 2), resolution))
if type == 's':
self.mixer = Residual(LocalWindowAttention(ed, kd, nh, attn_ratio=ar, \
resolution=resolution, window_resolution=window_resolution, kernels=kernels))
self.dw1 = Residual(Conv2d_BN(ed, ed, 3, 1, 1, groups=ed, bn_weight_init=0., resolution=resolution))
self.ffn1 = Residual(FFN(ed, int(ed * 2), resolution))
def forward(self, x):
return self.ffn1(self.dw1(self.mixer(self.ffn0(self.dw0(x)))))
class EfficientViT(torch.nn.Module):
def __init__(self, img_size=400,
patch_size=16,
frozen_stages=0,
in_chans=3,
stages=['s', 's', 's'],
embed_dim=[64, 128, 192],
key_dim=[16, 16, 16],
depth=[1, 2, 3],
num_heads=[4, 4, 4],
window_size=[7, 7, 7],
kernels=[5, 5, 5, 5],
down_ops=[['subsample', 2], ['subsample', 2], ['']],
pretrained=None,
distillation=False,):
super().__init__()
resolution = img_size
self.patch_embed = torch.nn.Sequential(Conv2d_BN(in_chans, embed_dim[0] // 8, 3, 2, 1, resolution=resolution), torch.nn.ReLU(),
Conv2d_BN(embed_dim[0] // 8, embed_dim[0] // 4, 3, 2, 1, resolution=resolution // 2), torch.nn.ReLU(),
Conv2d_BN(embed_dim[0] // 4, embed_dim[0] // 2, 3, 2, 1, resolution=resolution // 4), torch.nn.ReLU(),
Conv2d_BN(embed_dim[0] // 2, embed_dim[0], 3, 1, 1, resolution=resolution // 8))
resolution = img_size // patch_size
attn_ratio = [embed_dim[i] / (key_dim[i] * num_heads[i]) for i in range(len(embed_dim))]
self.blocks1 = []
self.blocks2 = []
self.blocks3 = []
for i, (stg, ed, kd, dpth, nh, ar, wd, do) in enumerate(
zip(stages, embed_dim, key_dim, depth, num_heads, attn_ratio, window_size, down_ops)):
for d in range(dpth):
eval('self.blocks' + str(i+1)).append(EfficientViTBlock(stg, ed, kd, nh, ar, resolution, wd, kernels))
if do[0] == 'subsample':
#('Subsample' stride)
blk = eval('self.blocks' + str(i+2))
resolution_ = (resolution - 1) // do[1] + 1
blk.append(torch.nn.Sequential(Residual(Conv2d_BN(embed_dim[i], embed_dim[i], 3, 1, 1, groups=embed_dim[i], resolution=resolution)),
Residual(FFN(embed_dim[i], int(embed_dim[i] * 2), resolution)),))
blk.append(PatchMerging(*embed_dim[i:i + 2], resolution))
resolution = resolution_
blk.append(torch.nn.Sequential(Residual(Conv2d_BN(embed_dim[i + 1], embed_dim[i + 1], 3, 1, 1, groups=embed_dim[i + 1], resolution=resolution)),
Residual(FFN(embed_dim[i + 1], int(embed_dim[i + 1] * 2), resolution)),))
self.blocks1 = torch.nn.Sequential(*self.blocks1)
self.blocks2 = torch.nn.Sequential(*self.blocks2)
self.blocks3 = torch.nn.Sequential(*self.blocks3)
self.channel = [i.size(1) for i in self.forward(torch.randn(1, 3, 640, 640))]
def forward(self, x):
outs = []
x = self.patch_embed(x)
x = self.blocks1(x)
outs.append(x)
x = self.blocks2(x)
outs.append(x)
x = self.blocks3(x)
outs.append(x)
return outs
EfficientViT_m0 = {
'img_size': 224,
'patch_size': 16,
'embed_dim': [64, 128, 192],
'depth': [1, 2, 3],
'num_heads': [4, 4, 4],
'window_size': [7, 7, 7],
'kernels': [7, 5, 3, 3],
}
EfficientViT_m1 = {
'img_size': 224,
'patch_size': 16,
'embed_dim': [128, 144, 192],
'depth': [1, 2, 3],
'num_heads': [2, 3, 3],
'window_size': [7, 7, 7],
'kernels': [7, 5, 3, 3],
}
EfficientViT_m2 = {
'img_size': 224,
'patch_size': 16,
'embed_dim': [128, 192, 224],
'depth': [1, 2, 3],
'num_heads': [4, 3, 2],
'window_size': [7, 7, 7],
'kernels': [7, 5, 3, 3],
}
EfficientViT_m3 = {
'img_size': 224,
'patch_size': 16,
'embed_dim': [128, 240, 320],
'depth': [1, 2, 3],
'num_heads': [4, 3, 4],
'window_size': [7, 7, 7],
'kernels': [5, 5, 5, 5],
}
EfficientViT_m4 = {
'img_size': 224,
'patch_size': 16,
'embed_dim': [128, 256, 384],
'depth': [1, 2, 3],
'num_heads': [4, 4, 4],
'window_size': [7, 7, 7],
'kernels': [7, 5, 3, 3],
}
EfficientViT_m5 = {
'img_size': 224,
'patch_size': 16,
'embed_dim': [192, 288, 384],
'depth': [1, 3, 4],
'num_heads': [3, 3, 4],
'window_size': [7, 7, 7],
'kernels': [7, 5, 3, 3],
}
def EfficientViT_M0(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m0):
model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)
if pretrained:
model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))
if fuse:
replace_batchnorm(model)
return model
def EfficientViT_M1(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m1):
model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)
if pretrained:
model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))
if fuse:
replace_batchnorm(model)
return model
def EfficientViT_M2(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m2):
model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)
if pretrained:
model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))
if fuse:
replace_batchnorm(model)
return model
def EfficientViT_M3(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m3):
model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)
if pretrained:
model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))
if fuse:
replace_batchnorm(model)
return model
def EfficientViT_M4(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m4):
model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)
if pretrained:
model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))
if fuse:
replace_batchnorm(model)
return model
def EfficientViT_M5(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m5):
model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)
if pretrained:
model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))
if fuse:
replace_batchnorm(model)
return model
def update_weight(model_dict, weight_dict):
idx, temp_dict = 0, {}
for k, v in weight_dict.items():
# k = k[9:]
if k in model_dict.keys() and np.shape(model_dict[k]) == np.shape(v):
temp_dict[k] = v
idx += 1
model_dict.update(temp_dict)
print(f'loading weights... {idx}/{len(model_dict)} items')
return model_dictyolov5-efficientvit.yaml
backbone:
# [from, number, module, args]
[[-1, 1, EfficientViT_M0, []], # 4
[-1, 1, SPPF, [1024, 5]], # 5
]
扫一扫
865
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
