Add RepVGG models ()

* Add RepVGG models

* Update model code

* Rename code

* format code

* add README of RepVGG
pull/595/head
jm12138 2021-02-01 18:07:25 +08:00 committed by GitHub
parent d08daf9b83
commit 8cd779be4d
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
17 changed files with 1254 additions and 0 deletions

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_A0'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_A1'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_A2'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B0'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B1'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B1g2'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B1g4'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B2'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B2g2'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B2g4'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B3'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B3g2'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,74 @@
mode: 'train'
ARCHITECTURE:
name: 'RepVGG_B3g4'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 120
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: -1
LEARNING_RATE:
function: 'Cosine'
params:
lr: 0.001
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.000100
TRAIN:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:

View File

@ -0,0 +1,27 @@
# RepVGG series
## Overview
RepVGG (Making VGG-style ConvNets Great Again) series model is a simple but powerful convolutional neural network architecture proposed by Tsinghua University (Guiguang Ding's team), MEGVII Technology (Jian Sun et al.), HKUST and Aberystwyth University in 2021. The architecture has an inference time agent similar to VGG. The main body is composed of 3x3 convolution and relu stack, while the training time model has multi branch topology. The decoupling of training time and inference time is realized by re-parameterization technology, so the model is called repvgg. [paper](https://arxiv.org/abs/2101.03697).
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1| FLOPS<br>(G) |
|:--:|:--:|:--:|:--:|:--:|
| RepVGG_A0 | 0.7131 | 0.9016 | 0.7241 | |
| RepVGG_A1 | 0.7380 | 0.9146 | 0.7446 | |
| RepVGG_A2 | 0.7571 | 0.9264 | 0.7648 | |
| RepVGG_B0 | 0.7450 | 0.9213 | 0.7514 | |
| RepVGG_B1 | 0.7773 | 0.9385 | 0.7837 | |
| RepVGG_B2 | 0.7813 | 0.9410 | 0.7878 | |
| RepVGG_B1g2 | 0.7732 | 0.9359 | 0.7778 | |
| RepVGG_B1g4 | 0.7675 | 0.9335 | 0.7758 | |
| RepVGG_B2g4 | 0.7782 | 0.9380 | 0.7850 | |
| Models | Top1 | Top5 | Reference<br>top1 | FLOPS<br>(G) |
|:--:|:--:|:--:|:--:|:--:|
| RepVGG_B3_200epochs | 0.7987 | 0.9502 | 0.8052 | |
| RepVGG_B2g4_200epochs | 0.7881 | 0.9448 | 0.7938 | |
| RepVGG_B3g4_200epochs | 0.7965 | 0.9485 | 0.8021 | |
Params, FLOPs, Inference speed and other information are coming soon.

View File

@ -0,0 +1,28 @@
# RepVGG系列
## 概述
RepVGGMaking VGG-style ConvNets Great Again系列模型是由清华大学(丁贵广团队)、旷视科技(孙剑等人)、港科大和阿伯里斯特威斯大学在2021年提出的一个简单但强大的卷积神经网络架构该架构具有类似于VGG的推理时间主体该主体仅由3x3卷积和ReLU的堆栈组成而训练时间模型具有多分支拓扑。训练时间和推理时间架构的这种解耦是通过结构重新参数化(re-parameterization)技术实现的因此该模型称为RepVGG。[论文地址](https://arxiv.org/abs/2101.03697)。
## 精度、FLOPS和参数量
| Models | Top1 | Top5 | Reference<br>top1| FLOPS<br>(G) |
|:--:|:--:|:--:|:--:|:--:|
| RepVGG_A0 | 0.7131 | 0.9016 | 0.7241 | |
| RepVGG_A1 | 0.7380 | 0.9146 | 0.7446 | |
| RepVGG_A2 | 0.7571 | 0.9264 | 0.7648 | |
| RepVGG_B0 | 0.7450 | 0.9213 | 0.7514 | |
| RepVGG_B1 | 0.7773 | 0.9385 | 0.7837 | |
| RepVGG_B2 | 0.7813 | 0.9410 | 0.7878 | |
| RepVGG_B1g2 | 0.7732 | 0.9359 | 0.7778 | |
| RepVGG_B1g4 | 0.7675 | 0.9335 | 0.7758 | |
| RepVGG_B2g4 | 0.7782 | 0.9380 | 0.7850 | |
| Models | Top1 | Top5 | Reference<br>top1 | FLOPS<br>(G) |
|:--:|:--:|:--:|:--:|:--:|
| RepVGG_B3_200epochs | 0.7987 | 0.9502 | 0.8052 | |
| RepVGG_B2g4_200epochs | 0.7881 | 0.9448 | 0.7938 | |
| RepVGG_B3g4_200epochs | 0.7965 | 0.9485 | 0.8021 | |
关于Params、FLOPs、Inference speed等信息敬请期待。

View File

@ -46,3 +46,5 @@ from .regnet import RegNetX_200MF, RegNetX_4GF, RegNetX_32GF, RegNetY_200MF, Reg
from .vision_transformer import ViT_small_patch16_224, ViT_base_patch16_224, ViT_base_patch16_384, ViT_base_patch32_384, ViT_large_patch16_224, ViT_large_patch16_384, ViT_large_patch32_384, ViT_huge_patch16_224, ViT_huge_patch32_384
from .distilled_vision_transformer import 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
from .distillation_models import ResNet50_vd_distill_MobileNetV3_large_x1_0
from .repvgg import RepVGG_A0, RepVGG_A1, RepVGG_A2, RepVGG_B0, RepVGG_B1, RepVGG_B2, RepVGG_B3, RepVGG_B1g2, RepVGG_B1g4, RepVGG_B2g2, RepVGG_B2g4, RepVGG_B3g2, RepVGG_B3g4

View File

@ -0,0 +1,235 @@
import paddle.nn as nn
import paddle
import numpy as np
__all__ = [
'RepVGG',
'RepVGG_A0', 'RepVGG_A1', 'RepVGG_A2',
'RepVGG_B0', 'RepVGG_B1', 'RepVGG_B2', 'RepVGG_B3',
'RepVGG_B1g2', 'RepVGG_B1g4',
'RepVGG_B2g2', 'RepVGG_B2g4',
'RepVGG_B3g2', 'RepVGG_B3g4',
]
class ConvBN(nn.Layer):
def __init__(self, in_channels, out_channels, kernel_size, stride, padding, groups=1):
super(ConvBN, self).__init__()
self.conv = nn.Conv2D(in_channels=in_channels, out_channels=out_channels,
kernel_size=kernel_size, stride=stride, padding=padding, groups=groups, bias_attr=False)
self.bn = nn.BatchNorm2D(num_features=out_channels)
def forward(self, x):
y = self.conv(x)
y = self.bn(y)
return y
class RepVGGBlock(nn.Layer):
def __init__(self, in_channels, out_channels, kernel_size,
stride=1, padding=0, dilation=1, groups=1, padding_mode='zeros'):
super(RepVGGBlock, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.padding = padding
self.dilation = dilation
self.groups = groups
self.padding_mode = padding_mode
assert kernel_size == 3
assert padding == 1
padding_11 = padding - kernel_size // 2
self.nonlinearity = nn.ReLU()
self.rbr_identity = nn.BatchNorm2D(
num_features=in_channels) if out_channels == in_channels and stride == 1 else None
self.rbr_dense = ConvBN(in_channels=in_channels, out_channels=out_channels,
kernel_size=kernel_size, stride=stride, padding=padding, groups=groups)
self.rbr_1x1 = ConvBN(in_channels=in_channels, out_channels=out_channels,
kernel_size=1, stride=stride, padding=padding_11, groups=groups)
def forward(self, inputs):
if not self.training:
return self.nonlinearity(self.rbr_reparam(inputs))
if self.rbr_identity is None:
id_out = 0
else:
id_out = self.rbr_identity(inputs)
return self.nonlinearity(self.rbr_dense(inputs) + self.rbr_1x1(inputs) + id_out)
def eval(self):
if not hasattr(self, 'rbr_reparam'):
self.rbr_reparam = nn.Conv2D(in_channels=self.in_channels, out_channels=self.out_channels, kernel_size=self.kernel_size, stride=self.stride,
padding=self.padding, dilation=self.dilation, groups=self.groups, padding_mode=self.padding_mode)
self.training = False
kernel, bias = self.get_equivalent_kernel_bias()
self.rbr_reparam.weight.set_value(kernel)
self.rbr_reparam.bias.set_value(bias)
for layer in self.sublayers():
layer.eval()
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 nn.functional.pad(kernel1x1, [1, 1, 1, 1])
def _fuse_bn_tensor(self, branch):
if branch is None:
return 0, 0
if isinstance(branch, ConvBN):
kernel = branch.conv.weight
running_mean = branch.bn._mean
running_var = branch.bn._variance
gamma = branch.bn.weight
beta = branch.bn.bias
eps = branch.bn._epsilon
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 = paddle.to_tensor(kernel_value)
kernel = self.id_tensor
running_mean = branch._mean
running_var = branch._variance
gamma = branch.weight
beta = branch.bias
eps = branch._epsilon
std = (running_var + eps).sqrt()
t = (gamma / std).reshape((-1, 1, 1, 1))
return kernel * t, beta - running_mean * gamma / std
class RepVGG(nn.Layer):
def __init__(self, num_blocks, width_multiplier=None, override_groups_map=None, class_dim=1000):
super(RepVGG, self).__init__()
assert len(width_multiplier) == 4
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]), class_dim)
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 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 = paddle.flatten(out, start_axis=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 RepVGG_A0(**kwargs):
return RepVGG(num_blocks=[2, 4, 14, 1],
width_multiplier=[0.75, 0.75, 0.75, 2.5], override_groups_map=None, **kwargs)
def RepVGG_A1(**kwargs):
return RepVGG(num_blocks=[2, 4, 14, 1],
width_multiplier=[1, 1, 1, 2.5], override_groups_map=None, **kwargs)
def RepVGG_A2(**kwargs):
return RepVGG(num_blocks=[2, 4, 14, 1],
width_multiplier=[1.5, 1.5, 1.5, 2.75], override_groups_map=None, **kwargs)
def RepVGG_B0(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[1, 1, 1, 2.5], override_groups_map=None, **kwargs)
def RepVGG_B1(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[2, 2, 2, 4], override_groups_map=None, **kwargs)
def RepVGG_B1g2(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[2, 2, 2, 4], override_groups_map=g2_map, **kwargs)
def RepVGG_B1g4(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[2, 2, 2, 4], override_groups_map=g4_map, **kwargs)
def RepVGG_B2(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=None, **kwargs)
def RepVGG_B2g2(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=g2_map, **kwargs)
def RepVGG_B2g4(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=g4_map, **kwargs)
def RepVGG_B3(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[3, 3, 3, 5], override_groups_map=None, **kwargs)
def RepVGG_B3g2(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[3, 3, 3, 5], override_groups_map=g2_map, **kwargs)
def RepVGG_B3g4(**kwargs):
return RepVGG(num_blocks=[4, 6, 16, 1],
width_multiplier=[3, 3, 3, 5], override_groups_map=g4_map, **kwargs)