paper/TA_EC
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update code for kd

Browse Source
This commit is contained in:
yoiannis 2025-03-09 22:36:22 +08:00
parent 4051332732
commit 49b2110b5f
9 changed files with 1295 additions and 55 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.gitignore vendored
View File

@ -1 +1,3 @@
data
*.pth
*.pyc

179
FED.py
View File

@ -6,45 +6,28 @@ from torch.utils.data import DataLoader, Subset
import numpy as np
import copy
from tqdm import tqdm
from model.repvit import repvit_m1_1
from model.mobilenetv3 import MobileNetV3
# 配置参数
NUM_CLIENTS = 10
NUM_CLIENTS = 4
NUM_ROUNDS = 3
CLIENT_EPOCHS = 2
CLIENT_EPOCHS = 5
BATCH_SIZE = 32
TEMP = 2.0 # 蒸馏温度
# 设备配置
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# 定义中心大模型
class ServerModel(nn.Module):
def __init__(self):
super().__init__()
self.fc1 = nn.Linear(784, 512)
self.fc2 = nn.Linear(512, 256)
self.fc3 = nn.Linear(256, 10)
def forward(self, x):
x = x.view(-1, 784)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
return self.fc3(x)
# 定义端侧小模型
class ClientModel(nn.Module):
def __init__(self):
super().__init__()
self.fc1 = nn.Linear(784, 64)
self.fc2 = nn.Linear(64, 10)
def forward(self, x):
x = x.view(-1, 784)
x = F.relu(self.fc1(x))
return self.fc2(x)
# 数据准备
def prepare_data(num_clients):
transform = transforms.Compose([transforms.ToTensor()])
transform = transforms.Compose([
transforms.Resize((224, 224)), # 将图像调整为 224x224
transforms.Grayscale(num_output_channels=3),
transforms.ToTensor()
])
train_set = datasets.MNIST("./data", train=True, download=True, transform=transform)
# 非IID数据划分（每个客户端2个类别）
@ -67,30 +50,80 @@ def client_train(client_model, server_model, dataset):
optimizer = torch.optim.SGD(client_model.parameters(), lr=0.1)
loader = DataLoader(dataset, batch_size=BATCH_SIZE, shuffle=True)
for _ in range(CLIENT_EPOCHS):
for data, target in loader:
# 训练进度条
progress_bar = tqdm(total=CLIENT_EPOCHS*len(loader),
desc="Client Training",
unit="batch")
for epoch in range(CLIENT_EPOCHS):
epoch_loss = 0.0
task_loss = 0.0
distill_loss = 0.0
correct = 0
total = 0
for batch_idx, (data, target) in enumerate(loader):
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
# 获取小模型输出
# 前向传播
client_output = client_model(data)
# 获取大模型输出（知识蒸馏）
# 获取教师模型输出
with torch.no_grad():
server_output = server_model(data)
# 计算联合损失
# 计算损失
loss_task = F.cross_entropy(client_output, target)
loss_distill = F.kl_div(
F.log_softmax(client_output/TEMP, dim=1),
F.softmax(server_output/TEMP, dim=1),
reduction="batchmean"
) * (TEMP**2)
total_loss = loss_task + loss_distill
# 反向传播
total_loss.backward()
optimizer.step()
# 统计指标
epoch_loss += total_loss.item()
task_loss += loss_task.item()
distill_loss += loss_distill.item()
_, predicted = torch.max(client_output.data, 1)
correct += (predicted == target).sum().item()
total += target.size(0)
# 实时更新进度条
progress_bar.set_postfix({
"Epoch": f"{epoch+1}/{CLIENT_EPOCHS}",
"Batch": f"{batch_idx+1}/{len(loader)}",
"Loss": f"{total_loss.item():.4f}",
"Acc": f"{100*correct/total:.2f}%\n",
})
progress_bar.update(1)
# 每10个batch打印详细信息
if (batch_idx + 1) % 10 == 0:
progress_bar.write(f"\nEpoch {epoch+1} | Batch {batch_idx+1}")
progress_bar.write(f"Task Loss: {loss_task:.4f}")
progress_bar.write(f"Distill Loss: {loss_distill:.4f}")
progress_bar.write(f"Total Loss: {total_loss:.4f}")
progress_bar.write(f"Batch Accuracy: {100*correct/total:.2f}%\n")
# 每个epoch结束打印汇总信息
avg_loss = epoch_loss / len(loader)
avg_task = task_loss / len(loader)
avg_distill = distill_loss / len(loader)
epoch_acc = 100 * correct / total
print(f"\n{'='*40}")
print(f"Epoch {epoch+1} Summary:")
print(f"Average Loss: {avg_loss:.4f}")
print(f"Task Loss: {avg_task:.4f}")
print(f"Distill Loss: {avg_distill:.4f}")
print(f"Training Accuracy: {epoch_acc:.2f}%")
print(f"{'='*40}\n")
progress_bar.close()
return client_model.state_dict()
# 模型参数聚合（FedAvg）
@ -105,7 +138,10 @@ def server_update(server_model, client_models, public_loader):
server_model.train()
optimizer = torch.optim.Adam(server_model.parameters(), lr=0.001)
for data, _ in public_loader:
total_loss = 0.0
progress_bar = tqdm(public_loader, desc="Server Updating", unit="batch")
for batch_idx, (data, _) in enumerate(progress_bar):
data = data.to(device)
optimizer.zero_grad()
@ -122,8 +158,19 @@ def server_update(server_model, client_models, public_loader):
reduction="batchmean"
)
# 反向传播
loss.backward()
optimizer.step()
# 更新统计信息
total_loss += loss.item()
progress_bar.set_postfix({
"Avg Loss": f"{total_loss/(batch_idx+1):.4f}",
"Current Loss": f"{loss.item():.4f}"
})
print(f"\nServer Update Complete | Average Loss: {total_loss/len(public_loader):.4f}\n")
def test_model(model, test_loader):
model.eval()
@ -139,36 +186,54 @@ def test_model(model, test_loader):
accuracy = 100 * correct / total
return accuracy
# 主训练流程
def main():
# 初始化模型
global_server_model = ServerModel().to(device)
client_models = [ClientModel().to(device) for _ in range(NUM_CLIENTS)]
global_server_model = repvit_m1_1(num_classes=10).to(device)
client_models = [MobileNetV3(n_class=10).to(device) for _ in range(NUM_CLIENTS)]
round_progress = tqdm(total=NUM_ROUNDS, desc="Federated Rounds", unit="round")
# 准备数据
client_datasets = prepare_data(NUM_CLIENTS)
public_loader = DataLoader(
datasets.MNIST("./data", train=False, download=True,
transform=transforms.ToTensor()),
transform= transforms.Compose([
transforms.Resize((224, 224)), # 将图像调整为 224x224
transforms.Grayscale(num_output_channels=3),
transforms.ToTensor() # 将图像转换为张量
])),
batch_size=100, shuffle=True)
test_dataset = datasets.MNIST(
"./data",
train=False,
transform= transforms.Compose([
transforms.Resize((224, 224)), # 将图像调整为 224x224
transforms.Grayscale(num_output_channels=3),
transforms.ToTensor() # 将图像转换为张量
])
)
test_loader = DataLoader(test_dataset, batch_size=100, shuffle=False)
for round in range(NUM_ROUNDS):
print(f"\n{'#'*50}")
print(f"Federated Round {round+1}/{NUM_ROUNDS}")
print(f"{'#'*50}")
# 客户端选择
selected_clients = np.random.choice(NUM_CLIENTS, 5, replace=False)
selected_clients = np.random.choice(NUM_CLIENTS, 2, replace=False)
print(f"Selected Clients: {selected_clients}")
# 客户端本地训练
client_params = []
for cid in selected_clients:
# 下载全局模型
print(f"\nTraining Client {cid}")
local_model = copy.deepcopy(client_models[cid])
local_model.load_state_dict(client_models[cid].state_dict())
# 本地训练
updated_params = client_train(
local_model,
global_server_model,
client_datasets[cid]
)
updated_params = client_train(local_model, global_server_model, client_datasets[cid])
client_params.append(updated_params)
# 模型聚合
@ -177,8 +242,18 @@ def main():
model.load_state_dict(global_client_params)
# 服务器知识更新
print("\nServer Updating...")
server_update(global_server_model, client_models, public_loader)
# 测试模型性能
server_acc = test_model(global_server_model, test_loader)
client_acc = test_model(client_models[0], test_loader)
print(f"\nRound {round+1} Performance:")
print(f"Global Model Accuracy: {server_acc:.2f}%")
print(f"Client Model Accuracy: {client_acc:.2f}%")
round_progress.update(1)
print(f"Round {round+1} completed")
print("Training completed!")
@ -189,21 +264,15 @@ def main():
print("Models saved successfully.")
# 创建测试数据加载器
test_dataset = datasets.MNIST(
"./data",
train=False,
transform=transforms.ToTensor()
)
test_loader = DataLoader(test_dataset, batch_size=100, shuffle=False)
# 测试服务器模型
server_model = ServerModel().to(device)
server_model = repvit_m1_1(num_classes=10).to(device)
server_model.load_state_dict(torch.load("server_model.pth"))
server_acc = test_model(server_model, test_loader)
print(f"Server Model Test Accuracy: {server_acc:.2f}%")
# 测试客户端模型
client_model = ClientModel().to(device)
client_model = MobileNetV3(n_class=10).to(device)
client_model.load_state_dict(torch.load("client_model.pth"))
client_acc = test_model(client_model, test_loader)
print(f"Client Model Test Accuracy: {client_acc:.2f}%")

157
dataset/recover.py Normal file
View File

@ -0,0 +1,157 @@
import random
import cv2
import numpy as np
import os
# 假设我们有一个函数 `remove_background` 使用某种方法去背景，返回前景掩码和前景图像
def remove_background(image_path):
# 加载图像
source_image = cv2.imread(image_path)
# 这里可以使用一个预训练的模型去背景，比如 U2-Net。为了简化，假设我们得到一个二值掩码
# 掩码生成逻辑可以替换为实际的模型推理
# 转换为灰度图像
GRAY = cv2.cvtColor(source_image, cv2.COLOR_BGR2GRAY)
# 二值化处理
_, mask_threshold = cv2.threshold(GRAY, 0, 1, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
# 定义结构元素
element = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5))
element1 = cv2.getStructuringElement(cv2.MORPH_RECT, (2, 2))
# 膨胀和腐蚀操作
mask_dilate = cv2.dilate(mask_threshold, element)
mask_erode = cv2.erode(mask_dilate, element1)
# 计算非零像素数量
count2 = cv2.countNonZero(mask_erode)
# 查找轮廓
contours, hierarchy = cv2.findContours(mask_erode, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_TC89_KCOS)
# 过滤轮廓
contours = [c for c in contours if cv2.contourArea(c) >= count2 * 0.3]
# 绘制轮廓
mask = np.zeros_like(mask_erode)
cv2.drawContours(mask, contours, -1, 1, -1)
# 将掩码转换为3通道
mask_cvtColor = cv2.cvtColor(mask, cv2.COLOR_GRAY2BGR)
# 应用掩码
source_image_multiply = cv2.multiply(source_image, mask_cvtColor)
# 转换为HSV颜色空间
imgHSV = cv2.cvtColor(source_image_multiply, cv2.COLOR_BGR2HSV)
# 定义HSV范围
scalarL = np.array([0, 46, 46])
scalarH = np.array([45, 255, 255])
# 根据HSV范围生成掩码
mask_inRange = cv2.inRange(imgHSV, scalarL, scalarH)
# 二值化处理
_, mask_tthreshold = cv2.threshold(mask_inRange, 0, 1, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
# 中值滤波
mask_medianBlur = cv2.medianBlur(mask_tthreshold, 7)
# 将掩码转换为3通道
mask_scvtColor = cv2.cvtColor(mask_medianBlur, cv2.COLOR_GRAY2BGR)
# 应用掩码
source_image = cv2.multiply(source_image, mask_scvtColor)
return source_image
def synthesize_background(foreground, background):
# 创建前景掩膜（非黑色区域）
gray = cv2.cvtColor(foreground, cv2.COLOR_BGR2GRAY)
_, mask = cv2.threshold(gray, 1, 255, cv2.THRESH_BINARY) # 阈值设为1以保留所有非纯黑像素
mask = cv2.GaussianBlur(mask, (5,5), 0) # 高斯模糊柔化边缘
_, mask = cv2.threshold(mask, 200, 255, cv2.THRESH_BINARY) # 重新二值化
# 精准形态学处理
kernel = np.ones((2,2), np.uint8)
mask = cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel, iterations=1) # 闭运算填充小孔
# 反转掩膜用于获取背景区域
mask_inv = cv2.bitwise_not(mask)
# 提取背景和前景的ROI区域
background_roi = cv2.bitwise_and(background, background, mask=mask_inv)
foreground_roi = cv2.bitwise_and(foreground, foreground, mask=mask)
# 合成图像
result = cv2.add(foreground_roi, background_roi)
return result
def edge_fill2(img):
(height, width, p) = img.shape
H = 2384
W = 1560
top = bottom=int((W - height) / 2)
left= right= int((H - width) / 2)
img_result = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=0)
return img_result
def process_images(input_folder, background_image_path, output_base):
"""
递归处理所有子文件夹并保持目录结构
"""
# 预处理背景路径（只需执行一次）
if os.path.isfile(background_image_path):
background_paths = [background_image_path]
else:
valid_ext = ['.jpg', '.jpeg', '.png', '.bmp', '.webp']
background_paths = [
os.path.join(background_image_path, f)
for f in os.listdir(background_image_path)
if os.path.splitext(f)[1].lower() in valid_ext
]
# 递归遍历输入目录
for root, dirs, files in os.walk(input_folder):
# 计算相对路径
relative_path = os.path.relpath(root, input_folder)
# 创建对应的输出目录
output_dir = os.path.join(output_base, relative_path)
os.makedirs(output_dir, exist_ok=True)
# 处理当前目录的文件
for filename in files:
input_path = os.path.join(root, filename)
output_path = os.path.join(output_dir, filename)
# 跳过非图像文件
if not filename.lower().endswith(('.png', '.jpg', '.jpeg', '.bmp')):
continue
try:
# 去背景处理
foreground = remove_background(input_path)
result = edge_fill2(foreground)
# 保存结果
cv2.imwrite(output_path, result)
print(f"Processed: {input_path} -> {output_path}")
except Exception as e:
print(f"Error processing {input_path}: {str(e)}")
# 使用示例
input_directory = 'L:/Tobacco/2023_JY/20230821/SOURCE'
background_image_path = 'F:/dataset/02.TA_EC/rundata/BACKGROUND/ZY_B'
output_directory = 'L:/Test'
process_images(input_directory, background_image_path, output_directory)

60
dataset/split.py Normal file
View File

@ -0,0 +1,60 @@
import os
import shutil
import random
def create_dataset_splits(base_dir, output_dir, train_ratio=0.7, val_ratio=0.15, test_ratio=0.15):
# 确保比例总和为1
assert train_ratio + val_ratio + test_ratio == 1.0, "Ratios must sum to 1"
# 创建输出目录
os.makedirs(output_dir, exist_ok=True)
train_dir = os.path.join(output_dir, 'train')
val_dir = os.path.join(output_dir, 'val')
test_dir = os.path.join(output_dir, 'test')
os.makedirs(train_dir, exist_ok=True)
os.makedirs(val_dir, exist_ok=True)
os.makedirs(test_dir, exist_ok=True)
categories = os.listdir(base_dir)
for category in categories:
category_path = os.path.join(base_dir, category)
if not os.path.isdir(category_path):
continue # 跳过非目录文件
# 获取类别文件夹中的所有图像文件
images = [f for f in os.listdir(category_path) if os.path.isfile(os.path.join(category_path, f))]
# 打乱图像文件顺序
random.shuffle(images)
# 计算分割点
total_images = len(images)
train_end = int(train_ratio * total_images)
val_end = train_end + int(val_ratio * total_images)
# 创建类别文件夹于各个输出目录中
os.makedirs(os.path.join(train_dir, category), exist_ok=True)
os.makedirs(os.path.join(val_dir, category), exist_ok=True)
os.makedirs(os.path.join(test_dir, category), exist_ok=True)
# 分配图像到训练集
for i in range(train_end):
shutil.copy(os.path.join(category_path, images[i]), os.path.join(train_dir, category))
# 分配图像到验证集
for i in range(train_end, val_end):
shutil.copy(os.path.join(category_path, images[i]), os.path.join(val_dir, category))
# 分配图像到测试集
for i in range(val_end, total_images):
shutil.copy(os.path.join(category_path, images[i]), os.path.join(test_dir, category))
print("Dataset successfully split into train, validation, and test sets.")
# 使用示例
base_directory = 'F:/dataset/02.TA_EC/EC27/JY_A'
output_directory = 'F:/dataset/02.TA_EC/datasets/EC27'
create_dataset_splits(base_directory, output_directory)

184
dataset/splitbc_compsition.py Normal file
View File

@ -0,0 +1,184 @@
import random
import cv2
import numpy as np
import os
# 假设我们有一个函数 `remove_background` 使用某种方法去背景，返回前景掩码和前景图像
def remove_background(image_path):
# 加载图像
source_image = cv2.imread(image_path)
# 这里可以使用一个预训练的模型去背景，比如 U2-Net。为了简化，假设我们得到一个二值掩码
# 掩码生成逻辑可以替换为实际的模型推理
# 转换为灰度图像
GRAY = cv2.cvtColor(source_image, cv2.COLOR_BGR2GRAY)
# 二值化处理
_, mask_threshold = cv2.threshold(GRAY, 0, 1, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
# 定义结构元素
element = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5))
element1 = cv2.getStructuringElement(cv2.MORPH_RECT, (2, 2))
# 膨胀和腐蚀操作
mask_dilate = cv2.dilate(mask_threshold, element)
mask_erode = cv2.erode(mask_dilate, element1)
# 计算非零像素数量
count2 = cv2.countNonZero(mask_erode)
# 查找轮廓
contours, hierarchy = cv2.findContours(mask_erode, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_TC89_KCOS)
# 过滤轮廓
contours = [c for c in contours if cv2.contourArea(c) >= count2 * 0.3]
# 绘制轮廓
mask = np.zeros_like(mask_erode)
cv2.drawContours(mask, contours, -1, 1, -1)
# 将掩码转换为3通道
mask_cvtColor = cv2.cvtColor(mask, cv2.COLOR_GRAY2BGR)
# 应用掩码
source_image_multiply = cv2.multiply(source_image, mask_cvtColor)
# 转换为HSV颜色空间
imgHSV = cv2.cvtColor(source_image_multiply, cv2.COLOR_BGR2HSV)
# 定义HSV范围
scalarL = np.array([0, 46, 46])
scalarH = np.array([45, 255, 255])
# 根据HSV范围生成掩码
mask_inRange = cv2.inRange(imgHSV, scalarL, scalarH)
# 二值化处理
_, mask_tthreshold = cv2.threshold(mask_inRange, 0, 1, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
# 中值滤波
mask_medianBlur = cv2.medianBlur(mask_tthreshold, 7)
# 将掩码转换为3通道
mask_scvtColor = cv2.cvtColor(mask_medianBlur, cv2.COLOR_GRAY2BGR)
# 应用掩码
source_image = cv2.multiply(source_image, mask_scvtColor)
return source_image
def synthesize_background(foreground, background_image_path):
# 验证输入路径有效性
if not os.path.exists(background_image_path):
raise FileNotFoundError(f"指定的背景路径不存在: {background_image_path}")
# 获取所有可用背景图像路径
background_paths = []
if os.path.isfile(background_image_path):
# 如果是单个图像文件
background_paths = [background_image_path]
elif os.path.isdir(background_image_path):
# 如果是目录，搜索常见图像格式
valid_extensions = ['.jpg', '.jpeg', '.png', '.bmp', '.webp']
for filename in os.listdir(background_image_path):
if os.path.splitext(filename)[1].lower() in valid_extensions:
background_paths.append(os.path.join(background_image_path, filename))
# 验证找到的图像文件数量
if not background_paths:
raise ValueError(f"目录中未找到支持的图像文件: {background_image_path}")
# 随机选择背景图像
selected_bg_path = random.choice(background_paths)
# 加载并验证背景图像
background = cv2.imread(selected_bg_path)
# 调整背景大小与前景一致
background = cv2.resize(background, (foreground.shape[1], foreground.shape[0]))
# 创建前景掩膜（非黑色区域）
gray = cv2.cvtColor(foreground, cv2.COLOR_BGR2GRAY)
_, mask = cv2.threshold(gray, 1, 255, cv2.THRESH_BINARY) # 阈值设为1以保留所有非纯黑像素
mask = cv2.GaussianBlur(mask, (5,5), 0) # 高斯模糊柔化边缘
_, mask = cv2.threshold(mask, 200, 255, cv2.THRESH_BINARY) # 重新二值化
# 精准形态学处理
kernel = np.ones((2,2), np.uint8)
mask = cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel, iterations=1) # 闭运算填充小孔
# 反转掩膜用于获取背景区域
mask_inv = cv2.bitwise_not(mask)
# 提取背景和前景的ROI区域
background_roi = cv2.bitwise_and(background, background, mask=mask_inv)
foreground_roi = cv2.bitwise_and(foreground, foreground, mask=mask)
# 合成图像
result = cv2.add(foreground_roi, background_roi)
# 保存结果
return result
def process_images(input_folder, background_image_path, output_base):
"""
递归处理所有子文件夹并保持目录结构
"""
# 预处理背景路径（只需执行一次）
if os.path.isfile(background_image_path):
background_paths = [background_image_path]
else:
valid_ext = ['.jpg', '.jpeg', '.png', '.bmp', '.webp']
background_paths = [
os.path.join(background_image_path, f)
for f in os.listdir(background_image_path)
if os.path.splitext(f)[1].lower() in valid_ext
]
# 递归遍历输入目录
for root, dirs, files in os.walk(input_folder):
# 计算相对路径
relative_path = os.path.relpath(root, input_folder)
# 创建对应的输出目录
output_dir = os.path.join(output_base, relative_path)
os.makedirs(output_dir, exist_ok=True)
# 处理当前目录的文件
for filename in files:
input_path = os.path.join(root, filename)
output_path = os.path.join(output_dir, filename)
# 跳过非图像文件
if not filename.lower().endswith(('.png', '.jpg', '.jpeg', '.bmp')):
continue
try:
# 去背景处理
foreground = remove_background(input_path)
# 随机选择背景（每次处理都重新选择）
bg_path = random.choice(background_paths)
background = cv2.imread(bg_path)
# 调整背景尺寸
h, w = foreground.shape[:2]
resized_bg = cv2.resize(background, (w, h))
# 合成背景
result = synthesize_background(foreground, bg_path)
# 保存结果
cv2.imwrite(output_path, result)
print(f"Processed: {input_path} -> {output_path}")
except Exception as e:
print(f"Error processing {input_path}: {str(e)}")
# 使用示例
input_directory = 'F:/dataset/02.TA_EC/EC27/JY_A/'
background_image_path = 'F:/dataset/02.TA_EC/rundata/BACKGROUND/ZY_B'
output_directory = 'F:/dataset/02.TA_EC/rundata/test'
process_images(input_directory, background_image_path, output_directory)

0
model/__init__.py Normal file
View File

248
model/mobilenetv3.py Normal file
View File

@ -0,0 +1,248 @@
import torch
import torch.nn as nn
import torch.nn.functional as F
__all__ = ['MobileNetV3', 'mobilenetv3']
def conv_bn(inp, oup, stride, conv_layer=nn.Conv2d, norm_layer=nn.BatchNorm2d, nlin_layer=nn.ReLU):
return nn.Sequential(
conv_layer(inp, oup, 3, stride, 1, bias=False),
norm_layer(oup),
nlin_layer(inplace=True)
)
def conv_1x1_bn(inp, oup, conv_layer=nn.Conv2d, norm_layer=nn.BatchNorm2d, nlin_layer=nn.ReLU):
return nn.Sequential(
conv_layer(inp, oup, 1, 1, 0, bias=False),
norm_layer(oup),
nlin_layer(inplace=True)
)
class Hswish(nn.Module):
def __init__(self, inplace=True):
super(Hswish, self).__init__()
self.inplace = inplace
def forward(self, x):
return x * F.relu6(x + 3., inplace=self.inplace) / 6.
class Hsigmoid(nn.Module):
def __init__(self, inplace=True):
super(Hsigmoid, self).__init__()
self.inplace = inplace
def forward(self, x):
return F.relu6(x + 3., inplace=self.inplace) / 6.
class SEModule(nn.Module):
def __init__(self, channel, reduction=4):
super(SEModule, self).__init__()
self.avg_pool = nn.AdaptiveAvgPool2d(1)
self.fc = nn.Sequential(
nn.Linear(channel, channel // reduction, bias=False),
nn.ReLU(inplace=True),
nn.Linear(channel // reduction, channel, bias=False),
Hsigmoid()
# nn.Sigmoid()
)
def forward(self, x):
b, c, _, _ = x.size()
y = self.avg_pool(x).view(b, c)
y = self.fc(y).view(b, c, 1, 1)
return x * y.expand_as(x)
class Identity(nn.Module):
def __init__(self, channel):
super(Identity, self).__init__()
def forward(self, x):
return x
def make_divisible(x, divisible_by=8):
import numpy as np
return int(np.ceil(x * 1. / divisible_by) * divisible_by)
class MobileBottleneck(nn.Module):
def __init__(self, inp, oup, kernel, stride, exp, se=False, nl='RE'):
super(MobileBottleneck, self).__init__()
assert stride in [1, 2]
assert kernel in [3, 5]
padding = (kernel - 1) // 2
self.use_res_connect = stride == 1 and inp == oup
conv_layer = nn.Conv2d
norm_layer = nn.BatchNorm2d
if nl == 'RE':
nlin_layer = nn.ReLU # or ReLU6
elif nl == 'HS':
nlin_layer = Hswish
else:
raise NotImplementedError
if se:
SELayer = SEModule
else:
SELayer = Identity
self.conv = nn.Sequential(
# pw
conv_layer(inp, exp, 1, 1, 0, bias=False),
norm_layer(exp),
nlin_layer(inplace=True),
# dw
conv_layer(exp, exp, kernel, stride, padding, groups=exp, bias=False),
norm_layer(exp),
SELayer(exp),
nlin_layer(inplace=True),
# pw-linear
conv_layer(exp, oup, 1, 1, 0, bias=False),
norm_layer(oup),
)
def forward(self, x):
if self.use_res_connect:
return x + self.conv(x)
else:
return self.conv(x)
class MobileNetV3(nn.Module):
def __init__(self, n_class=1000, input_size=224, dropout=0.8, mode='small', width_mult=1.0):
super(MobileNetV3, self).__init__()
input_channel = 16
last_channel = 1280
if mode == 'large':
# refer to Table 1 in paper
mobile_setting = [
# k, exp, c, se, nl, s,
[3, 16, 16, False, 'RE', 1],
[3, 64, 24, False, 'RE', 2],
[3, 72, 24, False, 'RE', 1],
[5, 72, 40, True, 'RE', 2],
[5, 120, 40, True, 'RE', 1],
[5, 120, 40, True, 'RE', 1],
[3, 240, 80, False, 'HS', 2],
[3, 200, 80, False, 'HS', 1],
[3, 184, 80, False, 'HS', 1],
[3, 184, 80, False, 'HS', 1],
[3, 480, 112, True, 'HS', 1],
[3, 672, 112, True, 'HS', 1],
[5, 672, 160, True, 'HS', 2],
[5, 960, 160, True, 'HS', 1],
[5, 960, 160, True, 'HS', 1],
]
elif mode == 'small':
# refer to Table 2 in paper
mobile_setting = [
# k, exp, c, se, nl, s,
[3, 16, 16, True, 'RE', 2],
[3, 72, 24, False, 'RE', 2],
[3, 88, 24, False, 'RE', 1],
[5, 96, 40, True, 'HS', 2],
[5, 240, 40, True, 'HS', 1],
[5, 240, 40, True, 'HS', 1],
[5, 120, 48, True, 'HS', 1],
[5, 144, 48, True, 'HS', 1],
[5, 288, 96, True, 'HS', 2],
[5, 576, 96, True, 'HS', 1],
[5, 576, 96, True, 'HS', 1],
]
else:
raise NotImplementedError
# building first layer
assert input_size % 32 == 0
last_channel = make_divisible(last_channel * width_mult) if width_mult > 1.0 else last_channel
self.features = [conv_bn(3, input_channel, 2, nlin_layer=Hswish)]
self.classifier = []
# building mobile blocks
for k, exp, c, se, nl, s in mobile_setting:
output_channel = make_divisible(c * width_mult)
exp_channel = make_divisible(exp * width_mult)
self.features.append(MobileBottleneck(input_channel, output_channel, k, s, exp_channel, se, nl))
input_channel = output_channel
# building last several layers
if mode == 'large':
last_conv = make_divisible(960 * width_mult)
self.features.append(conv_1x1_bn(input_channel, last_conv, nlin_layer=Hswish))
self.features.append(nn.AdaptiveAvgPool2d(1))
self.features.append(nn.Conv2d(last_conv, last_channel, 1, 1, 0))
self.features.append(Hswish(inplace=True))
elif mode == 'small':
last_conv = make_divisible(576 * width_mult)
self.features.append(conv_1x1_bn(input_channel, last_conv, nlin_layer=Hswish))
# self.features.append(SEModule(last_conv)) # refer to paper Table2, but I think this is a mistake
self.features.append(nn.AdaptiveAvgPool2d(1))
self.features.append(nn.Conv2d(last_conv, last_channel, 1, 1, 0))
self.features.append(Hswish(inplace=True))
else:
raise NotImplementedError
# make it nn.Sequential
self.features = nn.Sequential(*self.features)
# building classifier
self.classifier = nn.Sequential(
nn.Dropout(p=dropout), # refer to paper section 6
nn.Linear(last_channel, n_class),
)
self._initialize_weights()
def forward(self, x):
x = self.features(x)
x = x.mean(3).mean(2)
x = self.classifier(x)
return x
def _initialize_weights(self):
# weight initialization
for m in self.modules():
if isinstance(m, nn.Conv2d):
nn.init.kaiming_normal_(m.weight, mode='fan_out')
if m.bias is not None:
nn.init.zeros_(m.bias)
elif isinstance(m, nn.BatchNorm2d):
nn.init.ones_(m.weight)
nn.init.zeros_(m.bias)
elif isinstance(m, nn.Linear):
nn.init.normal_(m.weight, 0, 0.01)
if m.bias is not None:
nn.init.zeros_(m.bias)
def mobilenetv3(pretrained=False, **kwargs):
model = MobileNetV3(**kwargs)
if pretrained:
state_dict = torch.load('mobilenetv3_small_67.4.pth.tar')
model.load_state_dict(state_dict, strict=True)
# raise NotImplementedError
return model
if __name__ == '__main__':
net = mobilenetv3()
print('mobilenetv3:\n', net)
print('Total params: %.2fM' % (sum(p.numel() for p in net.parameters())/1000000.0))
input_size=(1, 3, 224, 224)
# pip install --upgrade git+https://github.com/kuan-wang/pytorch-OpCounter.git
from thop import profile
input_tensor = torch.randn(input_size)
flops, params = profile(net, inputs=(input_tensor,))
# print(flops)
# print(params)
print('Total params: %.2fM' % (params/1000000.0))
print('Total flops: %.2f GMACs' % ((flops/1000000000.0) / 2.0))
x = torch.randn(input_size)
out = net(x)

520
model/repvit.py Normal file
View File

@ -0,0 +1,520 @@
import torch.nn as nn
def _make_divisible(v, divisor, min_value=None):
"""
This function is taken from the original tf repo.
It ensures that all layers have a channel number that is divisible by 8
It can be seen here:
https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py
:param v:
:param divisor:
:param min_value:
:return:
"""
if min_value is None:
min_value = divisor
new_v = max(min_value, int(v + divisor / 2) // divisor * divisor)
# Make sure that round down does not go down by more than 10%.
if new_v < 0.9 * v:
new_v += divisor
return new_v
from timm.models.layers import SqueezeExcite
import torch
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,
device=c.weight.device)
m.weight.data.copy_(w)
m.bias.data.copy_(b)
return m
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)
@torch.no_grad()
def fuse(self):
if isinstance(self.m, Conv2d_BN):
m = self.m.fuse()
assert(m.groups == m.in_channels)
identity = torch.ones(m.weight.shape[0], m.weight.shape[1], 1, 1)
identity = torch.nn.functional.pad(identity, [1,1,1,1])
m.weight += identity.to(m.weight.device)
return m
elif isinstance(self.m, torch.nn.Conv2d):
m = self.m
assert(m.groups != m.in_channels)
identity = torch.ones(m.weight.shape[0], m.weight.shape[1], 1, 1)
identity = torch.nn.functional.pad(identity, [1,1,1,1])
m.weight += identity.to(m.weight.device)
return m
else:
return self
class RepVGGDW(torch.nn.Module):
def __init__(self, ed) -> None:
super().__init__()
self.conv = Conv2d_BN(ed, ed, 3, 1, 1, groups=ed)
self.conv1 = torch.nn.Conv2d(ed, ed, 1, 1, 0, groups=ed)
self.dim = ed
self.bn = torch.nn.BatchNorm2d(ed)
def forward(self, x):
return self.bn((self.conv(x) + self.conv1(x)) + x)
@torch.no_grad()
def fuse(self):
conv = self.conv.fuse()
conv1 = self.conv1
conv_w = conv.weight
conv_b = conv.bias
conv1_w = conv1.weight
conv1_b = conv1.bias
conv1_w = torch.nn.functional.pad(conv1_w, [1,1,1,1])
identity = torch.nn.functional.pad(torch.ones(conv1_w.shape[0], conv1_w.shape[1], 1, 1, device=conv1_w.device), [1,1,1,1])
final_conv_w = conv_w + conv1_w + identity
final_conv_b = conv_b + conv1_b
conv.weight.data.copy_(final_conv_w)
conv.bias.data.copy_(final_conv_b)
bn = self.bn
w = bn.weight / (bn.running_var + bn.eps)**0.5
w = conv.weight * w[:, None, None, None]
b = bn.bias + (conv.bias - bn.running_mean) * bn.weight / \
(bn.running_var + bn.eps)**0.5
conv.weight.data.copy_(w)
conv.bias.data.copy_(b)
return conv
class RepViTBlock(nn.Module):
def __init__(self, inp, hidden_dim, oup, kernel_size, stride, use_se, use_hs):
super(RepViTBlock, self).__init__()
assert stride in [1, 2]
self.identity = stride == 1 and inp == oup
assert(hidden_dim == 2 * inp)
if stride == 2:
self.token_mixer = nn.Sequential(
Conv2d_BN(inp, inp, kernel_size, stride, (kernel_size - 1) // 2, groups=inp),
SqueezeExcite(inp, 0.25) if use_se else nn.Identity(),
Conv2d_BN(inp, oup, ks=1, stride=1, pad=0)
)
self.channel_mixer = Residual(nn.Sequential(
# pw
Conv2d_BN(oup, 2 * oup, 1, 1, 0),
nn.GELU() if use_hs else nn.GELU(),
# pw-linear
Conv2d_BN(2 * oup, oup, 1, 1, 0, bn_weight_init=0),
))
else:
assert(self.identity)
self.token_mixer = nn.Sequential(
RepVGGDW(inp),
SqueezeExcite(inp, 0.25) if use_se else nn.Identity(),
)
self.channel_mixer = Residual(nn.Sequential(
# pw
Conv2d_BN(inp, hidden_dim, 1, 1, 0),
nn.GELU() if use_hs else nn.GELU(),
# pw-linear
Conv2d_BN(hidden_dim, oup, 1, 1, 0, bn_weight_init=0),
))
def forward(self, x):
return self.channel_mixer(self.token_mixer(x))
from timm.models.vision_transformer import trunc_normal_
class BN_Linear(torch.nn.Sequential):
def __init__(self, a, b, bias=True, std=0.02):
super().__init__()
self.add_module('bn', torch.nn.BatchNorm1d(a))
self.add_module('l', torch.nn.Linear(a, b, bias=bias))
trunc_normal_(self.l.weight, std=std)
if bias:
torch.nn.init.constant_(self.l.bias, 0)
@torch.no_grad()
def fuse(self):
bn, l = self._modules.values()
w = bn.weight / (bn.running_var + bn.eps)**0.5
b = bn.bias - self.bn.running_mean * \
self.bn.weight / (bn.running_var + bn.eps)**0.5
w = l.weight * w[None, :]
if l.bias is None:
b = b @ self.l.weight.T
else:
b = (l.weight @ b[:, None]).view(-1) + self.l.bias
m = torch.nn.Linear(w.size(1), w.size(0), device=l.weight.device)
m.weight.data.copy_(w)
m.bias.data.copy_(b)
return m
class Classfier(nn.Module):
def __init__(self, dim, num_classes, distillation=True):
super().__init__()
self.classifier = BN_Linear(dim, num_classes) if num_classes > 0 else torch.nn.Identity()
self.distillation = distillation
if distillation:
self.classifier_dist = BN_Linear(dim, num_classes) if num_classes > 0 else torch.nn.Identity()
def forward(self, x):
if self.distillation:
x = self.classifier(x), self.classifier_dist(x)
if not self.training:
x = (x[0] + x[1]) / 2
else:
x = self.classifier(x)
return x
@torch.no_grad()
def fuse(self):
classifier = self.classifier.fuse()
if self.distillation:
classifier_dist = self.classifier_dist.fuse()
classifier.weight += classifier_dist.weight
classifier.bias += classifier_dist.bias
classifier.weight /= 2
classifier.bias /= 2
return classifier
else:
return classifier
class RepViT(nn.Module):
def __init__(self, cfgs, num_classes=1000, distillation=False):
super(RepViT, self).__init__()
# setting of inverted residual blocks
self.cfgs = cfgs
# building first layer
input_channel = self.cfgs[0][2]
patch_embed = torch.nn.Sequential(Conv2d_BN(3, input_channel // 2, 3, 2, 1), torch.nn.GELU(),
Conv2d_BN(input_channel // 2, input_channel, 3, 2, 1))
layers = [patch_embed]
# building inverted residual blocks
block = RepViTBlock
for k, t, c, use_se, use_hs, s in self.cfgs:
output_channel = _make_divisible(c, 8)
exp_size = _make_divisible(input_channel * t, 8)
layers.append(block(input_channel, exp_size, output_channel, k, s, use_se, use_hs))
input_channel = output_channel
self.features = nn.ModuleList(layers)
self.classifier = Classfier(output_channel, num_classes, distillation)
def forward(self, x):
# x = self.features(x)
for f in self.features:
x = f(x)
x = torch.nn.functional.adaptive_avg_pool2d(x, 1).flatten(1)
x = self.classifier(x)
return x
from timm.models import register_model
def repvit_m0_6(pretrained=False, num_classes = 1000, distillation=False):
"""
Constructs a MobileNetV3-Large model
"""
cfgs = [
[3, 2, 40, 1, 0, 1],
[3, 2, 40, 0, 0, 1],
[3, 2, 80, 0, 0, 2],
[3, 2, 80, 1, 0, 1],
[3, 2, 80, 0, 0, 1],
[3, 2, 160, 0, 1, 2],
[3, 2, 160, 1, 1, 1],
[3, 2, 160, 0, 1, 1],
[3, 2, 160, 1, 1, 1],
[3, 2, 160, 0, 1, 1],
[3, 2, 160, 1, 1, 1],
[3, 2, 160, 0, 1, 1],
[3, 2, 160, 1, 1, 1],
[3, 2, 160, 0, 1, 1],
[3, 2, 160, 0, 1, 1],
[3, 2, 320, 0, 1, 2],
[3, 2, 320, 1, 1, 1],
]
return RepViT(cfgs, num_classes=num_classes, distillation=distillation)
def repvit_m0_9(pretrained=False, num_classes = 1000, distillation=False):
"""
Constructs a MobileNetV3-Large model
"""
cfgs = [
# k, t, c, SE, HS, s
[3, 2, 48, 1, 0, 1],
[3, 2, 48, 0, 0, 1],
[3, 2, 48, 0, 0, 1],
[3, 2, 96, 0, 0, 2],
[3, 2, 96, 1, 0, 1],
[3, 2, 96, 0, 0, 1],
[3, 2, 96, 0, 0, 1],
[3, 2, 192, 0, 1, 2],
[3, 2, 192, 1, 1, 1],
[3, 2, 192, 0, 1, 1],
[3, 2, 192, 1, 1, 1],
[3, 2, 192, 0, 1, 1],
[3, 2, 192, 1, 1, 1],
[3, 2, 192, 0, 1, 1],
[3, 2, 192, 1, 1, 1],
[3, 2, 192, 0, 1, 1],
[3, 2, 192, 1, 1, 1],
[3, 2, 192, 0, 1, 1],
[3, 2, 192, 1, 1, 1],
[3, 2, 192, 0, 1, 1],
[3, 2, 192, 1, 1, 1],
[3, 2, 192, 0, 1, 1],
[3, 2, 192, 0, 1, 1],
[3, 2, 384, 0, 1, 2],
[3, 2, 384, 1, 1, 1],
[3, 2, 384, 0, 1, 1]
]
return RepViT(cfgs, num_classes=num_classes, distillation=distillation)
def repvit_m1_0(pretrained=False, num_classes = 1000, distillation=False):
"""
Constructs a MobileNetV3-Large model
"""
cfgs = [
# k, t, c, SE, HS, s
[3, 2, 56, 1, 0, 1],
[3, 2, 56, 0, 0, 1],
[3, 2, 56, 0, 0, 1],
[3, 2, 112, 0, 0, 2],
[3, 2, 112, 1, 0, 1],
[3, 2, 112, 0, 0, 1],
[3, 2, 112, 0, 0, 1],
[3, 2, 224, 0, 1, 2],
[3, 2, 224, 1, 1, 1],
[3, 2, 224, 0, 1, 1],
[3, 2, 224, 1, 1, 1],
[3, 2, 224, 0, 1, 1],
[3, 2, 224, 1, 1, 1],
[3, 2, 224, 0, 1, 1],
[3, 2, 224, 1, 1, 1],
[3, 2, 224, 0, 1, 1],
[3, 2, 224, 1, 1, 1],
[3, 2, 224, 0, 1, 1],
[3, 2, 224, 1, 1, 1],
[3, 2, 224, 0, 1, 1],
[3, 2, 224, 1, 1, 1],
[3, 2, 224, 0, 1, 1],
[3, 2, 224, 0, 1, 1],
[3, 2, 448, 0, 1, 2],
[3, 2, 448, 1, 1, 1],
[3, 2, 448, 0, 1, 1]
]
return RepViT(cfgs, num_classes=num_classes, distillation=distillation)
def repvit_m1_1(pretrained=False, num_classes = 1000, distillation=False):
"""
Constructs a MobileNetV3-Large model
"""
cfgs = [
# k, t, c, SE, HS, s
[3, 2, 64, 1, 0, 1],
[3, 2, 64, 0, 0, 1],
[3, 2, 64, 0, 0, 1],
[3, 2, 128, 0, 0, 2],
[3, 2, 128, 1, 0, 1],
[3, 2, 128, 0, 0, 1],
[3, 2, 128, 0, 0, 1],
[3, 2, 256, 0, 1, 2],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 512, 0, 1, 2],
[3, 2, 512, 1, 1, 1],
[3, 2, 512, 0, 1, 1]
]
return RepViT(cfgs, num_classes=num_classes, distillation=distillation)
def repvit_m1_5(pretrained=False, num_classes = 1000, distillation=False):
"""
Constructs a MobileNetV3-Large model
"""
cfgs = [
# k, t, c, SE, HS, s
[3, 2, 64, 1, 0, 1],
[3, 2, 64, 0, 0, 1],
[3, 2, 64, 1, 0, 1],
[3, 2, 64, 0, 0, 1],
[3, 2, 64, 0, 0, 1],
[3, 2, 128, 0, 0, 2],
[3, 2, 128, 1, 0, 1],
[3, 2, 128, 0, 0, 1],
[3, 2, 128, 1, 0, 1],
[3, 2, 128, 0, 0, 1],
[3, 2, 128, 0, 0, 1],
[3, 2, 256, 0, 1, 2],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 1, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 256, 0, 1, 1],
[3, 2, 512, 0, 1, 2],
[3, 2, 512, 1, 1, 1],
[3, 2, 512, 0, 1, 1],
[3, 2, 512, 1, 1, 1],
[3, 2, 512, 0, 1, 1]
]
return RepViT(cfgs, num_classes=num_classes, distillation=distillation)
def repvit_m2_3(pretrained=False, num_classes = 1000, distillation=False):
"""
Constructs a MobileNetV3-Large model
"""
cfgs = [
# k, t, c, SE, HS, s
[3, 2, 80, 1, 0, 1],
[3, 2, 80, 0, 0, 1],
[3, 2, 80, 1, 0, 1],
[3, 2, 80, 0, 0, 1],
[3, 2, 80, 1, 0, 1],
[3, 2, 80, 0, 0, 1],
[3, 2, 80, 0, 0, 1],
[3, 2, 160, 0, 0, 2],
[3, 2, 160, 1, 0, 1],
[3, 2, 160, 0, 0, 1],
[3, 2, 160, 1, 0, 1],
[3, 2, 160, 0, 0, 1],
[3, 2, 160, 1, 0, 1],
[3, 2, 160, 0, 0, 1],
[3, 2, 160, 0, 0, 1],
[3, 2, 320, 0, 1, 2],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 320, 1, 1, 1],
[3, 2, 320, 0, 1, 1],
# [3, 2, 320, 1, 1, 1],
# [3, 2, 320, 0, 1, 1],
[3, 2, 320, 0, 1, 1],
[3, 2, 640, 0, 1, 2],
[3, 2, 640, 1, 1, 1],
[3, 2, 640, 0, 1, 1],
# [3, 2, 640, 1, 1, 1],
# [3, 2, 640, 0, 1, 1]
]
return RepViT(cfgs, num_classes=num_classes, distillation=distillation)
if __name__ == '__main__':
net = repvit_m1_1()
print('mobilenetv3:\n', net)
print('Total params: %.2fM' % (sum(p.numel() for p in net.parameters())/1000000.0))
input_size=(4, 3, 224, 224)
# pip install --upgrade git+https://github.com/kuan-wang/pytorch-OpCounter.git
from thop import profile
input_tensor = torch.randn(input_size)
flops, params = profile(net, inputs=(input_tensor,))
# print(flops)
# print(params)
print('Total params: %.2fM' % (params/1000000.0))
print('Total flops: %.2f GMACs' % ((flops/1000000000.0) / 2.0))
x = torch.randn(input_size)
out = net(x)

0
main.py → test.py
View File