點擊上方“AI算法與圖像處理”,選擇加"星標"或“置頂”
重磅乾貨,第一時間送達
極市導讀
本文將介紹一個優秀的PyTorch開源庫——timm庫,並對其中的vision transformer.py代碼進行了詳細解讀。
什麼是timm庫?
PyTorchImageModels,簡稱timm,是一個巨大的PyTorch代碼集合,包括了一系列:
-
image models -
layers -
utilities -
optimizers -
schedulers -
data-loaders / augmentations -
training / validation scripts
旨在將各種SOTA模型整合在一起,並具有復現ImageNet訓練結果的能力。
timm庫作者是來自加拿大溫哥華的Ross Wightman。
作者github鏈接:
https://github.com/rwightman
timm庫鏈接:
https://github.com/rwightman/pytorch-image-models
所有的PyTorch模型及其對應arxiv鏈接如下:
timm庫特點
所有的模型都有默認的API:
-
accessing/changing the classifier - get_classifier
andreset_classifier
-
只對features做前向傳播 - forward_features
所有模型都支持多尺度特徵提取 (feature pyramids) (通過create_model函數):
-
create_model(name, features_only=True, out_indices=..., output_stride=...)
out_indices
指定返回哪個feature maps to return, 從0開始,out_indices[i]
對應着 C(i + 1)
feature level。
output_stride
通過dilated convolutions控制網絡的output stride。大多數網絡默認 stride 32 。
所有的模型都有一致的pretrained weight loader,adapts last linear if necessary。
訓練方式支持:
-
NVIDIA DDP w/ a single GPU per process, multiple processes with APEX present (AMP mixed-precision optional) -
PyTorch DistributedDataParallel w/ multi-gpu, single process (AMP disabled as it crashes when enabled) -
PyTorch w/ single GPU single process (AMP optional)
動態的全局池化方式可以選擇: average pooling, max pooling, average + max, or concat([average, max]),默認是adaptive average。
Schedulers:
Schedulers 包括step
,cosine
w/ restarts,tanh
w/ restarts,plateau
。
Optimizer:
-
rmsprop_tf
adapted from PyTorch RMSProp by myself. Reproduces much improved Tensorflow RMSProp behaviour. -
radam
by Liyuan Liu (https://arxiv.org/abs/1908.03265) -
novograd
by Masashi Kimura (https://arxiv.org/abs/1905.11286) -
lookahead
adapted from impl by Liam (https://arxiv.org/abs/1907.08610) -
fused<name>
optimizers by name with NVIDIA Apex installed -
adamp
andsgdp
by Naver ClovAI (https://arxiv.org/abs/2006.08217) -
adafactor
adapted from FAIRSeq impl (https://arxiv.org/abs/1804.04235) -
adahessian
by David Samuel (https://arxiv.org/abs/2006.00719)
timm庫 vision_transformer.py代碼解讀
代碼來自:
https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py
對應的論文是ViT,是除了官方開源的代碼之外的又一個優秀的PyTorch implement。
An Image Is Worth 16 x 16 Words: Transformers for Image Recognition at Scale
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
https://arxiv.org/abs/2010.11929
另一篇工作DeiT也大量借鑑了timm庫這份代碼的實現:
Training data-efficient image transformers & distillation through attention
Training data-efficient image transformers & distillation through attention
https://arxiv.org/abs/2012.12877
vision_transformer.py:
代碼中定義的變量的含義如下:
img_size:tuple 類型,裏面是int類型,代表輸入的圖片大小,默認是 224。
patch_size:tuple 類型,裏面是int類型,代表Patch的大小,默認是 16。
in_chans:int 類型,代表輸入圖片的channel數,默認是3。
num_classes:int 類型classification head的分類數,比如CIFAR100就是100,默認是 1000。
embed_dim:int 類型Transformer的embedding dimension,默認是 768。
depth:int 類型,Transformer的Block的數量,默認是 12。
num_heads:int 類型,attention heads的數量,默認是12。
mlp_ratio:int 類型,mlp hidden dim/embedding dim的值,默認是 4。
qkv_bias:bool 類型,attention模塊計算qkv時需要bias嗎,默認是 True。
qk_scale: 一般設置成 None 就行。
drop_rate:float 類型,dropout rate,默認是 0。
attn_drop_rate:float 類型,attention模塊的dropout rate,默認是 0。
drop_path_rate:float 類型,默認是 0。
hybrid_backbone:nn.Module 類型,在把圖片轉換成Patch之前,需要先通過一個Backbone嗎?默認是 None。
如果是None,就直接把圖片轉化成Patch。
如果不是None,就先通過這個Backbone,再轉化成Patch。
norm_layer:nn.Module 類型,歸一化層類型,默認是 None。
1. 導入必要的庫和模型:
import math
import logging
from functools import partial
from collections import OrderedDict
import torch
import torch.nn as nn
import torch.nn.functional as F
from timm.data import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD
from .helpers import load_pretrained
from .layers import StdConv2dSame, DropPath, to_2tuple, trunc_normal_
from .resnet import resnet26d, resnet50d
from .resnetv2 import ResNetV2
from .registry import register_model
2. 定義一個字典,代表標準的模型,如果需要更改模型超參數只需要改變_cfg
的傳入的參數即可。
def _cfg(url='', **kwargs):
return {
'url': url,
'num_classes': 1000, 'input_size': (3, 224, 224), 'pool_size': None,
'crop_pct': .9, 'interpolation': 'bicubic',
'mean': IMAGENET_DEFAULT_MEAN, 'std': IMAGENET_DEFAULT_STD,
'first_conv': 'patch_embed.proj', 'classifier': 'head',
**kwargs
}
3. default_cfgs代表支持的所有模型,也定義成字典的形式:
vit_small_patch16_224裏面的small代表小模型。
ViT的第一步要把圖片分成一個個patch,然後把這些patch組合在一起作爲對圖像的序列化操作,比如一張224 × 224的圖片分成大小爲16 × 16的patch,那一共可以分成196個。所以這個圖片就序列化成了(196, 256)的tensor。所以這裏的:
16: 就代表patch的大小。
224: 就代表輸入圖片的大小。
按照這個命名方式,支持的模型有:vit_base_patch16_224,vit_base_patch16_384等等。後面的vit_deit_base_patch16_224等等模型代表DeiT這篇論文的模型。
default_cfgs = {
# patch models (my experiments)
'vit_small_patch16_224': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/vit_small_p16_224-15ec54c9.pth',
),
# patch models (weights ported from official Google JAX impl)
'vit_base_patch16_224': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_p16_224-80ecf9dd.pth',
mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5),
),
'vit_base_patch32_224': _cfg(
url='', # no official model weights for this combo, only for in21k
mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
'vit_base_patch16_384': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_p16_384-83fb41ba.pth',
input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0),
'vit_base_patch32_384': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_p32_384-830016f5.pth',
input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0),
'vit_large_patch16_224': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_p16_224-4ee7a4dc.pth',
mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
'vit_large_patch32_224': _cfg(
url='', # no official model weights for this combo, only for in21k
mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
'vit_large_patch16_384': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_p16_384-b3be5167.pth',
input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0),
'vit_large_patch32_384': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_p32_384-9b920ba8.pth',
input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0),
# patch models, imagenet21k (weights ported from official Google JAX impl)
'vit_base_patch16_224_in21k': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_patch16_224_in21k-e5005f0a.pth',
num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
'vit_base_patch32_224_in21k': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_patch32_224_in21k-8db57226.pth',
num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
'vit_large_patch16_224_in21k': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_patch16_224_in21k-606da67d.pth',
num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
'vit_large_patch32_224_in21k': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_patch32_224_in21k-9046d2e7.pth',
num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
'vit_huge_patch14_224_in21k': _cfg(
url='', # FIXME I have weights for this but > 2GB limit for github release binaries
num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
# hybrid models (weights ported from official Google JAX impl)
'vit_base_resnet50_224_in21k': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_resnet50_224_in21k-6f7c7740.pth',
num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=0.9, first_conv='patch_embed.backbone.stem.conv'),
'vit_base_resnet50_384': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_resnet50_384-9fd3c705.pth',
input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0, first_conv='patch_embed.backbone.stem.conv'),
# hybrid models (my experiments)
'vit_small_resnet26d_224': _cfg(),
'vit_small_resnet50d_s3_224': _cfg(),
'vit_base_resnet26d_224': _cfg(),
'vit_base_resnet50d_224': _cfg(),
# deit models (FB weights)
'vit_deit_tiny_patch16_224': _cfg(
url='https://dl.fbaipublicfiles.com/deit/deit_tiny_patch16_224-a1311bcf.pth'),
'vit_deit_small_patch16_224': _cfg(
url='https://dl.fbaipublicfiles.com/deit/deit_small_patch16_224-cd65a155.pth'),
'vit_deit_base_patch16_224': _cfg(
url='https://dl.fbaipublicfiles.com/deit/deit_base_patch16_224-b5f2ef4d.pth',),
'vit_deit_base_patch16_384': _cfg(
url='https://dl.fbaipublicfiles.com/deit/deit_base_patch16_384-8de9b5d1.pth',
input_size=(3, 384, 384), crop_pct=1.0),
'vit_deit_tiny_distilled_patch16_224': _cfg(
url='https://dl.fbaipublicfiles.com/deit/deit_tiny_distilled_patch16_224-b40b3cf7.pth'),
'vit_deit_small_distilled_patch16_224': _cfg(
url='https://dl.fbaipublicfiles.com/deit/deit_small_distilled_patch16_224-649709d9.pth'),
'vit_deit_base_distilled_patch16_224': _cfg(
url='https://dl.fbaipublicfiles.com/deit/deit_base_distilled_patch16_224-df68dfff.pth', ),
'vit_deit_base_distilled_patch16_384': _cfg(
url='https://dl.fbaipublicfiles.com/deit/deit_base_distilled_patch16_384-d0272ac0.pth',
input_size=(3, 384, 384), crop_pct=1.0),
}
4. FFN實現:
class Mlp(nn.Module):
def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
super().__init__()
out_features = out_features or in_features
hidden_features = hidden_features or in_features
self.fc1 = nn.Linear(in_features, hidden_features)
self.act = act_layer()
self.fc2 = nn.Linear(hidden_features, out_features)
self.drop = nn.Dropout(drop)
def forward(self, x):
x = self.fc1(x)
x = self.act(x)
x = self.drop(x)
x = self.fc2(x)
x = self.drop(x)
return x
5. Attention實現:
在python 3.5以後,@是一個操作符,表示矩陣-向量乘法
A@x 就是矩陣-向量乘法A*x: np.dot(A, x)。
class Attention(nn.Module):
def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0.):
super().__init__()
self.num_heads = num_heads
head_dim = dim // num_heads
self.scale = qk_scale or head_dim ** -0.5
self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
self.attn_drop = nn.Dropout(attn_drop)
self.proj = nn.Linear(dim, dim)
self.proj_drop = nn.Dropout(proj_drop)
def forward(self, x):
B, N, C = x.shape
qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
q, k, v = qkv[0], qkv[1], qkv[2]
attn = (q @ k.transpose(-2, -1)) * self.scale
attn = attn.softmax(dim=-1)
attn = self.attn_drop(attn)
x = (attn @ v).transpose(1, 2).reshape(B, N, C)
x = self.proj(x)
x = self.proj_drop(x)
return x
6. 包含Attention和Add & Norm的Block實現:
圖1:Block類對應結構
不同之處是:
先進行Norm,再Attention;先進行Norm,再通過FFN (MLP)。
class Block(nn.Module):
def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,
drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm):
super().__init__()
self.norm1 = norm_layer(dim)
self.attn = Attention(
dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)
# NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()
self.norm2 = norm_layer(dim)
mlp_hidden_dim = int(dim * mlp_ratio)
self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)
def forward(self, x):
x = x + self.drop_path(self.attn(self.norm1(x)))
x = x + self.drop_path(self.mlp(self.norm2(x)))
return x
7. 接下來要把圖片轉換成Patch,一種做法是直接把Image轉化成Patch,另一種做法是把Backbone輸出的特徵轉化成Patch。
1) 直接把Image轉化成Patch:
輸入的x的維度是:(B, C, H, W)
輸出的PatchEmbedding的維度是:(B, 14*14, 768),768表示embed_dim,14*14表示一共有196個Patches。
class PatchEmbed(nn.Module):
""" Image to Patch Embedding
"""
def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768):
super().__init__()
img_size = to_2tuple(img_size)
patch_size = to_2tuple(patch_size)
num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0])
self.img_size = img_size
self.patch_size = patch_size
self.num_patches = num_patches
self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
def forward(self, x):
B, C, H, W = x.shape
# FIXME look at relaxing size constraints
assert H == self.img_size[0] and W == self.img_size[1], \
f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})."
x = self.proj(x).flatten(2).transpose(1, 2)
# x: (B, 14*14, 768)
return x
2) 把Backbone輸出的特徵轉化成Patch:
輸入的x的維度是:(B, C, H, W)
得到Backbone輸出的維度是:(B, feature_size, feature_size, feature_dim)
輸出的PatchEmbedding的維度是:(B, feature_size, feature_size, embed_dim),一共有feature_size * feature_size個Patches。
class HybridEmbed(nn.Module):
""" CNN Feature Map Embedding
Extract feature map from CNN, flatten, project to embedding dim.
"""
def __init__(self, backbone, img_size=224, feature_size=None, in_chans=3, embed_dim=768):
super().__init__()
assert isinstance(backbone, nn.Module)
img_size = to_2tuple(img_size)
self.img_size = img_size
self.backbone = backbone
if feature_size is None:
with torch.no_grad():
# FIXME this is hacky, but most reliable way of determining the exact dim of the output feature
# map for all networks, the feature metadata has reliable channel and stride info, but using
# stride to calc feature dim requires info about padding of each stage that isn't captured.
training = backbone.training
if training:
backbone.eval()
o = self.backbone(torch.zeros(1, in_chans, img_size[0], img_size[1]))
if isinstance(o, (list, tuple)):
o = o[-1] # last feature if backbone outputs list/tuple of features
feature_size = o.shape[-2:]
feature_dim = o.shape[1]
backbone.train(training)
else:
feature_size = to_2tuple(feature_size)
if hasattr(self.backbone, 'feature_info'):
feature_dim = self.backbone.feature_info.channels()[-1]
else:
feature_dim = self.backbone.num_features
self.num_patches = feature_size[0] * feature_size[1]
self.proj = nn.Conv2d(feature_dim, embed_dim, 1)
def forward(self, x):
x = self.backbone(x)
if isinstance(x, (list, tuple)):
x = x[-1] # last feature if backbone outputs list/tuple of features
x = self.proj(x).flatten(2).transpose(1, 2)
return x
8. 以上是ViT所需的所有模塊的定義,下面是VisionTransformer 這個類的實現:
8.1 使用這個類時需要傳入的變量,其含義已經在本小節一開始介紹。
class VisionTransformer(nn.Module):
""" Vision Transformer
A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale` -
https://arxiv.org/abs/2010.11929
"""
def __init__(self, img_size=224, patch_size=16, in_chans=3, num_classes=1000, embed_dim=768, depth=12,
num_heads=12, mlp_ratio=4., qkv_bias=True, qk_scale=None, representation_size=None,
drop_rate=0., attn_drop_rate=0., drop_path_rate=0., hybrid_backbone=None, norm_layer=None):
8.2 得到分塊後的Patch的數量:
super().__init__()
self.num_classes = num_classes
self.num_features = self.embed_dim = embed_dim # num_features for consistency with other models
norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6)
if hybrid_backbone is not None:
self.patch_embed = HybridEmbed(
hybrid_backbone, img_size=img_size, in_chans=in_chans, embed_dim=embed_dim)
else:
self.patch_embed = PatchEmbed(
img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim)
num_patches = self.patch_embed.num_patches
8.3 class token:
一開始定義成(1, 1, 768),之後再變成(B, 1, 768)。
self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
8.4 定義位置編碼:
self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim))
8.5 把12個Block連接起來:
self.pos_drop = nn.Dropout(p=drop_rate)
dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]
self.blocks = nn.ModuleList([
Block(
dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale,
drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[i], norm_layer=norm_layer)
for i in range(depth)])
self.norm = norm_layer(embed_dim)
8.6 表示層和分類頭:
表示層輸出維度是representation_size,分類頭輸出維度是num_classes。
# Representation layer
if representation_size:
self.num_features = representation_size
self.pre_logits = nn.Sequential(OrderedDict([
('fc', nn.Linear(embed_dim, representation_size)),
('act', nn.Tanh())
]))
else:
self.pre_logits = nn.Identity()
# Classifier head
self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity()
8.7 初始化各個模塊:
函數trunc_normal_(tensor, mean=0., std=1., a=-2., b=2.)的目的是用截斷的正態分佈繪製的值填充輸入張量,我們只需要輸入均值mean,標準差std,下界a,上界b即可。
self.apply(self._init_weights)表示對各個模塊的權重進行初始化。apply函數的代碼是:
for module in self.children():
module.apply(fn)
fn(self)
return self
遞歸地將fn應用於每個子模塊,相當於在遞歸調用fn,即_init_weights這個函數。
也就是把模型的所有子模塊的nn.Linear和nn.LayerNorm層都初始化掉。
trunc_normal_(self.pos_embed, std=.02)
trunc_normal_(self.cls_token, std=.02)
self.apply(self._init_weights)
def _init_weights(self, m):
if isinstance(m, nn.Linear):
trunc_normal_(m.weight, std=.02)
if isinstance(m, nn.Linear) and m.bias is not None:
nn.init.constant_(m.bias, 0)
elif isinstance(m, nn.LayerNorm):
nn.init.constant_(m.bias, 0)
nn.init.constant_(m.weight, 1.0)
8.8 最後就是整個ViT模型的forward實現:
def forward_features(self, x):
B = x.shape[0]
x = self.patch_embed(x)
cls_tokens = self.cls_token.expand(B, -1, -1) # stole cls_tokens impl from Phil Wang, thanks
x = torch.cat((cls_tokens, x), dim=1)
x = x + self.pos_embed
x = self.pos_drop(x)
for blk in self.blocks:
x = blk(x)
x = self.norm(x)[:, 0]
x = self.pre_logits(x)
return x
def forward(self, x):
x = self.forward_features(x)
x = self.head(x)
return x
9. 下面是Training data-efficient image transformers & distillation through attention這篇論文的DeiT這個類的實現:
整體結構與ViT相似,繼承了上面的VisionTransformer類。
class DistilledVisionTransformer(VisionTransformer):
再額外定義以下3個變量:
-
distillation token:dist_token -
新的位置編碼:pos_embed -
蒸餾分類頭:head_dist
DeiT相關介紹可以參考:Vision Transformer 超詳細解讀 (原理分析+代碼解讀) (三)。
self.dist_token = nn.Parameter(torch.zeros(1, 1, self.embed_dim))
num_patches = self.patch_embed.num_patches
self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 2, self.embed_dim))
self.head_dist = nn.Linear(self.embed_dim, self.num_classes) if self.num_classes > 0 else nn.Identity()
初始化新定義的變量:
trunc_normal_(self.dist_token, std=.02)
trunc_normal_(self.pos_embed, std=.02)
self.head_dist.apply(self._init_weights)
前向函數:
def forward_features(self, x):
B = x.shape[0]
x = self.patch_embed(x)
cls_tokens = self.cls_token.expand(B, -1, -1) # stole cls_tokens impl from Phil Wang, thanks
dist_token = self.dist_token.expand(B, -1, -1)
x = torch.cat((cls_tokens, dist_token, x), dim=1)
x = x + self.pos_embed
x = self.pos_drop(x)
for blk in self.blocks:
x = blk(x)
x = self.norm(x)
return x[:, 0], x[:, 1]
def forward(self, x):
x, x_dist = self.forward_features(x)
x = self.head(x)
x_dist = self.head_dist(x_dist)
if self.training:
return x, x_dist
else:
# during inference, return the average of both classifier predictions
return (x + x_dist) / 2
10. 對位置編碼進行插值:
posemb代表未插值的位置編碼權值,posemb_tok爲位置編碼的token部分,posemb_grid爲位置編碼的插值部分。
首先把要插值部分posemb_grid給reshape成(1, gs_old, gs_old, -1)的形式,再插值成(1, gs_new, gs_new, -1)的形式,最後與token部分在第1維度拼接在一起,得到插值後的位置編碼posemb。
def resize_pos_embed(posemb, posemb_new):
# Rescale the grid of position embeddings when loading from state_dict. Adapted from
# https://github.com/google-research/vision_transformer/blob/00883dd691c63a6830751563748663526e811cee/vit_jax/checkpoint.py#L224
position embedding: %s to %s', posemb.shape, posemb_new.shape)
ntok_new = posemb_new.shape[1]
if True:
posemb_grid = posemb[:, :1], posemb[0, 1:]
ntok_new -= 1
else:
posemb_grid = posemb[:, :0], posemb[0]
gs_old = int(math.sqrt(len(posemb_grid)))
gs_new = int(math.sqrt(ntok_new))
embedding grid-size from %s to %s', gs_old, gs_new)
posemb_grid = posemb_grid.reshape(1, gs_old, gs_old, -1).permute(0, 3, 1, 2)
posemb_grid = F.interpolate(posemb_grid, size=(gs_new, gs_new), mode='bilinear')
posemb_grid = posemb_grid.permute(0, 2, 3, 1).reshape(1, gs_new * gs_new, -1)
posemb = torch.cat([posemb_tok, posemb_grid], dim=1)
return posemb
11. _create_vision_transformer函數用於創建vision transformer:
checkpoint_filter_fn的作用是加載預訓練權重。
def checkpoint_filter_fn(state_dict, model):
""" convert patch embedding weight from manual patchify + linear proj to conv"""
out_dict = {}
if 'model' in state_dict:
# For deit models
state_dict = state_dict['model']
for k, v in state_dict.items():
if 'patch_embed.proj.weight' in k and len(v.shape) < 4:
# For old models that I trained prior to conv based patchification
O, I, H, W = model.patch_embed.proj.weight.shape
v = v.reshape(O, -1, H, W)
elif k == 'pos_embed' and v.shape != model.pos_embed.shape:
# To resize pos embedding when using model at different size from pretrained weights
v = resize_pos_embed(v, model.pos_embed)
out_dict[k] = v
return out_dict
def _create_vision_transformer(variant, pretrained=False, distilled=False, **kwargs):
default_cfg = default_cfgs[variant]
default_num_classes = default_cfg['num_classes']
default_img_size = default_cfg['input_size'][-1]
num_classes = kwargs.pop('num_classes', default_num_classes)
img_size = kwargs.pop('img_size', default_img_size)
repr_size = kwargs.pop('representation_size', None)
if repr_size is not None and num_classes != default_num_classes:
# Remove representation layer if fine-tuning. This may not always be the desired action,
# but I feel better than doing nothing by default for fine-tuning. Perhaps a better interface?
_logger.warning("Removing representation layer for fine-tuning.")
repr_size = None
model_cls = DistilledVisionTransformer if distilled else VisionTransformer
model = model_cls(img_size=img_size, num_classes=num_classes, representation_size=repr_size, **kwargs)
model.default_cfg = default_cfg
if pretrained:
load_pretrained(
model, num_classes=num_classes, in_chans=kwargs.get('in_chans', 3),
filter_fn=partial(checkpoint_filter_fn, model=model))
return model
12. 定義和註冊vision transformer模型:
@ 指裝飾器。
@register_model代表註冊器,註冊這個新定義的模型。
model_kwargs是一個存有模型所有超參數的字典。
最後使用上面定義的_create_vision_transformer函數創建模型。
def vit_base_patch16_224(pretrained=False, **kwargs):
""" ViT-Base (ViT-B/16) from original paper (https://arxiv.org/abs/2010.11929).
ImageNet-1k weights fine-tuned from in21k @ 224x224, source https://github.com/google-research/vision_transformer.
"""
model_kwargs = dict(patch_size=16, embed_dim=768, depth=12, num_heads=12, **kwargs)
model = _create_vision_transformer('vit_base_patch16_224', pretrained=pretrained, **model_kwargs)
return model
一共可以選擇的模型包括:
ViT系列:
vit_small_patch16_224
vit_base_patch16_224
vit_base_patch32_224
vit_base_patch16_384
vit_base_patch32_384
vit_large_patch16_224
vit_large_patch32_224
vit_large_patch16_384
vit_large_patch32_384
vit_base_patch16_224_in21k
vit_base_patch32_224_in21k
vit_large_patch16_224_in21k
vit_large_patch32_224_in21k
vit_huge_patch14_224_in21k
vit_base_resnet50_224_in21k
vit_base_resnet50_384
vit_small_resnet26d_224
vit_small_resnet50d_s3_224
vit_base_resnet26d_224
vit_base_resnet50d_224DeiT系列:
vit_deit_tiny_patch16_224
vit_deit_small_patch16_224
vit_deit_base_patch16_224
vit_deit_base_patch16_384
vit_deit_tiny_distilled_patch16_224
vit_deit_small_distilled_patch16_224
vit_deit_base_distilled_patch16_224
vit_deit_base_distilled_patch16_384
以上就是對timm庫 vision_transformer.py代碼的分析。
如何使用timm庫以及 vision_transformer.py代碼搭建自己的模型?
在搭建我們自己的視覺Transformer模型時,我們可以按照下面的步驟操作:首先
-
繼承timm庫的 VisionTransformer這個類。 -
添加上自己模型 獨有的一些變量。 -
重寫 forward函數。 -
通過timm庫的 註冊器註冊新模型。
我們以ViT模型的改進版DeiT爲例:
首先,DeiT的所有模型列表如下:
__all__ = [
'deit_tiny_patch16_224', 'deit_small_patch16_224', 'deit_base_patch16_224',
'deit_tiny_distilled_patch16_224', 'deit_small_distilled_patch16_224',
'deit_base_distilled_patch16_224', 'deit_base_patch16_384',
'deit_base_distilled_patch16_384',
]
導入VisionTransformer這個類,註冊器register_model,以及初始化函數trunc_normal_:
from timm.models.vision_transformer import VisionTransformer, _cfg
from timm.models.registry import register_model
from timm.models.layers import trunc_normal_
DeiT的class名稱是DistilledVisionTransformer,它直接繼承了VisionTransformer這個類:
class DistilledVisionTransformer(VisionTransformer):
添加上自己模型獨有的一些變量:
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.dist_token = nn.Parameter(torch.zeros(1, 1, self.embed_dim))
num_patches = self.patch_embed.num_patches
# 位置編碼不是ViT中的(b, N, 256), 而變成了(b, N+2, 256), 原因是還有class token和distillation token.
self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 2, self.embed_dim))
self.head_dist = nn.Linear(self.embed_dim, self.num_classes) if self.num_classes > 0 else nn.Identity()
trunc_normal_(self.dist_token, std=.02)
trunc_normal_(self.pos_embed, std=.02)
self.head_dist.apply(self._init_weights)
重寫forward函數:
def forward_features(self, x):
# taken from https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py
# with slight modifications to add the dist_token
B = x.shape[0]
x = self.patch_embed(x)
cls_tokens = self.cls_token.expand(B, -1, -1) # stole cls_tokens impl from Phil Wang, thanks
dist_token = self.dist_token.expand(B, -1, -1)
x = torch.cat((cls_tokens, dist_token, x), dim=1)
x = x + self.pos_embed
x = self.pos_drop(x)
for blk in self.blocks:
x = blk(x)
x = self.norm(x)
return x[:, 0], x[:, 1]
def forward(self, x):
x, x_dist = self.forward_features(x)
x = self.head(x)
x_dist = self.head_dist(x_dist)
if self.training:
return x, x_dist
else:
# during inference, return the average of both classifier predictions
return (x + x_dist) / 2
通過timm庫的註冊器註冊新模型:
def deit_base_patch16_224(pretrained=False, **kwargs):
model = VisionTransformer(
patch_size=16, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4, qkv_bias=True,
norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
model.default_cfg = _cfg()
if pretrained:
checkpoint = torch.hub.load_state_dict_from_url(
url="https://dl.fbaipublicfiles.com/deit/deit_base_patch16_224-b5f2ef4d.pth",
map_location="cpu", check_hash=True
)
model.load_state_dict(checkpoint["model"])
return model
個人微信(如果沒有備註不拉羣!)
請註明:
地區+學校/企業+研究方向+暱稱
下載1:何愷明頂會分享
在「AI算法與圖像處理」公衆號後臺回覆:何愷明,即可下載。總共有6份PDF,涉及 ResNet、Mask RCNN等經典工作的總結分析
下載2:終身受益的編程指南:Google編程風格指南
在「AI算法與圖像處理」公衆號後臺回覆:c++,即可下載。歷經十年考驗,最權威的編程規範!
下載3 CVPR2020
在「AI算法與圖像處理」公衆號後臺回覆:
CVPR2020
,即可下載1467篇CVPR 2020論文
覺得不錯就點亮在看吧
本文分享自微信公衆號 - AI算法與圖像處理(AI_study)。
如有侵權,請聯繫 [email protected] 刪除。
本文參與“OSC源創計劃”,歡迎正在閱讀的你也加入,一起分享。