ViT Vision Transformer超详细解析，网络构建，可视化，数据预处理，全流程实例教程

关于ViT的分析和教程，网上又虚又空的东西比较多，本文通过一个实例，将ViT全解析。

包括三部分内容，网络构建；orchview.draw_graph 将网络每一层的结构与输入输出可视化；数据预处理。附完整代码

网络构建

创建一个model.py，其中实现ViT网络构建

import torch.nn as nn
import torch
import torch.optim as optim
import torch.nn.functional as F
import lightning as L


class AttentionBlock(nn.Module):
    def __init__(self, embed_dim, hidden_dim, num_heads, dropout=0.0):
        """
        Inputs:
            embed_dim - Dimensionality of input and attention feature vectors
            hidden_dim - Dimensionality of hidden layer in feed-forward network
                         (usually 2-4x larger than embed_dim)
            num_heads - Number of heads to use in the Multi-Head Attention block
            dropout - Amount of dropout to apply in the feed-forward network
        """
        super().__init__()

        self.layer_norm_1 = nn.LayerNorm(embed_dim)
        self.attn = nn.MultiheadAttention(embed_dim, num_heads)
        self.layer_norm_2 = nn.LayerNorm(embed_dim)
        self.linear = nn.Sequential(
            nn.Linear(embed_dim, hidden_dim),
            nn.GELU(),
            nn.Dropout(dropout),
            nn.Linear(hidden_dim, embed_dim),
            nn.Dropout(dropout),
        )

    def forward(self, x):
        inp_x = self.layer_norm_1(x)
        x = x + self.attn(inp_x, inp_x, inp_x)[0]
        x = x + self.linear(self.layer_norm_2(x))
        return x


class VisionTransformer(nn.Module):
    def __init__(
        self,
        embed_dim,
        hidden_dim,
        num_channels,
        num_heads,
        num_layers,
        num_classes,
        patch_size,
        num_patches,
        dropout=0.0,
    ):
        """
        Inputs:
            embed_dim - Dimensionality of the input feature vectors to the Transformer
            hidden_dim - Dimensionality of the hidden layer in the feed-forward networks
                         within the Transformer
            num_channels - Number of channels of the input (3 for RGB)
            num_heads - Number of heads to use in the Multi-Head Attention block
            num_layers - Number of layers to use in the Transformer
            num_classes - Number of classes to predict
            patch_size - Number of pixels that the patches have per dimension
            num_patches - Maximum number of patches an image can have
            dropout - Amount of dropout to apply in the feed-forward network and
                      on the input encoding
        """
        super().__init__()

        self.patch_size = patch_size

        # Layers/Networks
        self.input_layer = nn.Linear(num_channels * (patch_size**2), embed_dim)
        self.transformer = nn.Sequential(
            *(AttentionBlock(embed_dim, hidden_dim, num_heads, dropout=dropout) for _ in range(num_layers))
        )
        self.mlp_head = nn.Sequential(nn.LayerNorm(embed_dim), nn.Linear(embed_dim, num_classes))
        self.dropout = nn.Dropout(dropout)

        # Parameters/Embeddings
        self.cls_token = nn.Parameter(torch.randn(1, 1, embed_dim))
        self.pos_embedding = nn.Parameter(torch.randn(1, 1 + num_patches, embed_dim))

    def img_to_patch(self, x, patch_size, flatten_channels=True):
        """
        Inputs:
            x - Tensor representing the image of shape [B, C, H, W]
            patch_size - Number of pixels per dimension of the patches (integer)
            flatten_channels - If True, the patches will be returned in a flattened format
                               as a feature vector instead of a image grid.
        """
        B, C, H, W = x.shape
        x = x.reshape(B, C, H // patch_size, patch_size, W // patch_size, patch_size)
        x = x.permute(0, 2, 4, 1, 3, 5)  # [B, H', W', C, p_H, p_W]
        x = x.flatten(1, 2)  # [B, H'*W', C, p_H, p_W]
        if flatten_channels:
            x = x.flatten(2, 4)  # [B, H'*W', C*p_H*p_W]
        return x

    def forward(self, x):
        # Preprocess input
        x = self.img_to_patch(x, self.patch_size)
        B, T, _ = x.shape
        x = self.input_layer(x)

        # Add CLS token and positional encoding
        cls_token = self.cls_token.repeat(B, 1, 1)
        x = torch.cat([cls_token, x], dim=1)
        x = x + self.pos_embedding[:, : T + 1]

        # Apply Transforrmer
        x = self.dropout(x)
        x = x.transpose(0, 1)
        x = self.transformer(x)

        # Perform classification prediction
        cls = x[0]
        out = self.mlp_head(cls)
        return out


class ViT(L.LightningModule):
    def __init__(self, model_kwargs, lr):
        super().__init__()
        self.save_hyperparameters()
        self.model = VisionTransformer(**model_kwargs)

    def forward(self, x):
        return self.model(x)

    def configure_optimizers(self):
        optimizer = optim.AdamW(self.parameters(), lr=self.hparams.lr)
        lr_scheduler = optim.lr_scheduler.MultiStepLR(optimizer, milestones=[100, 150], gamma=0.1)
        return [optimizer], [lr_scheduler]

    def _calculate_loss(self, batch, mode="train"):
        imgs, labels = batch
        preds = self.model(imgs)
        loss = F.cross_entropy(preds, labels)
        acc = (preds.argmax(dim=-1) == labels).float().mean()

        self.log("%s_loss" % mode, loss)
        self.log("%s_acc" % mode, acc)
        return loss

    def training_step(self, batch, batch_idx):
        loss = self._calculate_loss(batch, mode="train")
        return loss

    def validation_step(self, batch, batch_idx):
        self._calculate_loss(batch, mode="val")

    def test_step(self, batch, batch_idx):
        self._calculate_loss(batch, mode="test")

在其他文件中引入model.py，实现网络搭建

from model import ViT

model = ViT(model_kwargs={
        "embed_dim": 256,
        "hidden_dim": 512,
        "num_heads": 8,
        "num_layers": 6,
        "patch_size": 4,
        "num_channels": 3,
        "num_patches": 64,
        "num_classes": 10,
        "dropout": 0.2,
        },
        lr=3e-4,
    )

也可以下载预训练的模型

# Files to download
base_url = "https://raw.githubusercontent.com/phlippe/saved_models/main/"
CHECKPOINT_PATH = os.environ.get("PATH_CHECKPOINT", "saved_models/VisionTransformers/")
pretrained_files = [
    "tutorial15/ViT.ckpt",
    "tutorial15/tensorboards/ViT/events.out.tfevents.ViT",
    "tutorial5/tensorboards/ResNet/events.out.tfevents.resnet",
]
# Create checkpoint path if it doesn't exist yet
os.makedirs(CHECKPOINT_PATH, exist_ok=True)

# For each file, check whether it already exists. If not, try downloading it.
for file_name in pretrained_files:
    file_path = os.path.join(CHECKPOINT_PATH, file_name.split("/", 1)[1])
    if "/" in file_name.split("/", 1)[1]:
        os.makedirs(file_path.rsplit("/", 1)[0], exist_ok=True)
    if not os.path.isfile(file_path):
        file_url = base_url + file_name
        print("Downloading %s..." % file_url)
        try:
            urllib.request.urlretrieve(file_url, file_path)
        except HTTPError as e:
            print(
                "Something went wrong. Please try to download the file from the GDrive folder, or contact the author with the full output including the following error:\n",
                e,
            )

pretrained_filename = os.path.join(CHECKPOINT_PATH, "ViT.ckpt")
model = ViT.load_from_checkpoint(pretrained_filename)

torchview.draw_graph 网络可视化

model_graph = draw_graph(model, input_size=(1, 3, 16, 16))
model_graph.resize_graph(scale=5.0)
model_graph.visual_graph.render(format='svg')

运行这段代码，会生成一个svg格式的图片，显示网络结构和每一层的输入输出

训练数据准备

新建一个prepare_data.py

import os
import json
import torch
from PIL import Image
from torch.utils.data import Dataset, DataLoader
import torchvision.transforms as transforms


class CustomDataset(Dataset):
    def __init__(self, image_dir, names, labels, transform=None):
        self.image_dir = image_dir
        self.names = names
        self.labels = labels
        self.transform = transform

    def __len__(self):
        return len(self.labels)

    def __getitem__(self, idx):
        name_ = self.names[idx]
        img_name = os.path.join(self.image_dir, name_)
        image = Image.open(img_name)

        if self.transform:
            image = self.transform(image)

        label = self.labels[idx]

        return image, label


def load_img_ann(ann_path):
    """return [{img_name, [ (x, y, h, w, label), ... ]}]"""
    with open(ann_path) as fp:
        root = json.load(fp)
    img_dict = {}
    img_label_dict = {}
    for img_info in root['images']:
        img_id = img_info['id']
        img_name = img_info['file_name']
        img_dict[img_id] = {'name': img_name}
    for ann_info in root['annotations']:
        img_id = ann_info['image_id']
        img_category_id = ann_info['category_id']
        img_name = img_dict[img_id]['name']

        img_label_dict[img_id] = {'name': img_name, 'category_id': img_category_id}

    return img_label_dict


def get_dataloader():
    annota_dir = '/home/robotics/Downloads/coco_dataset/annotations/instances_val2017.json'
    img_dir = "/home/robotics/Downloads/coco_dataset/val2017"
    img_dict = load_img_ann(annota_dir)
    values = list(img_dict.values())
    img_names = []
    labels = []
    for item in values:
        category_id = item['category_id']
        labels.append(category_id)
        img_name = item['name']
        img_names.append(img_name)

    # 检查剔除黑白的图片
    img_names_rgb = []
    labels_rgb = []
    for i in range(len(img_names)):
        # 检查文件扩展名，确保它是图片文件（可以根据需要扩展支持的文件类型）
        file_path = os.path.join(img_dir, img_names[i])

        # 打开图片文件
        img = Image.open(file_path)

        # 获取通道数
        num_channels = img.mode
        if num_channels == "RGB" and labels[i] < 10:
            img_names_rgb.append(img_names[i])
            labels_rgb.append(labels[i])

    # 定义一系列图像转换操作
    transform = transforms.Compose([
        transforms.Resize((16, 16)),  # 调整图像大小
        transforms.RandomHorizontalFlip(),  # 随机水平翻转
        transforms.ToTensor(),  # 将图像转换为张量
        transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])  # 标准化图像
    ])

    # 假设 image_dir 是包含所有图像文件的文件夹路径，labels 是标签列表
    train_set = CustomDataset(img_dir, img_names_rgb[-500:], labels_rgb[-500:], transform=transform)
    val_set = CustomDataset(img_dir, img_names_rgb[-500:-100], labels_rgb[-500:-100], transform=transform)
    test_set = CustomDataset(img_dir, img_names_rgb[-100:], labels_rgb[-100:], transform=transform)

    # 创建一个 DataLoader
    train_loader = DataLoader(train_set, batch_size=32, shuffle=True, drop_last=False)
    val_loader = DataLoader(val_set, batch_size=32, shuffle=True, drop_last=False, num_workers=4)
    test_loader = DataLoader(test_set, batch_size=32, shuffle=True, drop_last=False, num_workers=4)

    return train_loader, val_loader, test_loader


if __name__ == "__main__":
    train_loader, val_loader, test_loader = get_dataloader()
    for batch in train_loader:
        print(batch)

解释一下上面的代码：

这里使用的是coco数据集的2017，可以在官网自行下载，下载下来以后，annotations包含如下内容

这里我们使用的是 instances_val2017.json，如果是正经做训练，应该用train2017，但是train2017文件太大了，处理起来速度很慢，本文仅为说明，不追求训练效果，所以使用val2017进行说明，instances就是用于图像识别的annotation，里面包括了每张图片的label和box，本文创建的ViT 不输出box，仅输出类别。函数

def load_img_ann(ann_path):

是为了将图片的id(每张图片的唯一主键)，name和category_id(属于哪一个类别，也就是label)关联起来。

        # 获取通道数
        num_channels = img.mode
        if num_channels == "RGB" and labels[i] < 10:
            img_names_rgb.append(img_names[i])
            labels_rgb.append(labels[i])

注意coco数据集有单通道的黑白图片，要剔除，因为本文的ViT比较简单，输出只能10个类别，所以预处理图片的时候，只选择10个类别。

定义操作变换

    # 定义一系列图像转换操作
    transform = transforms.Compose([
        transforms.Resize((16, 16)),  # 调整图像大小
        transforms.RandomHorizontalFlip(),  # 随机水平翻转
        transforms.ToTensor(),  # 将图像转换为张量
        transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])  # 标准化图像
    ])

创建一个自己的Dataset类，继承自 torch.utils.data.Dataset

class CustomDataset(Dataset):
    def __init__(self, image_dir, names, labels, transform=None):
        self.image_dir = image_dir
        self.names = names
        self.labels = labels
        self.transform = transform

    def __len__(self):
        return len(self.labels)

    def __getitem__(self, idx):
        name_ = self.names[idx]
        img_name = os.path.join(self.image_dir, name_)
        image = Image.open(img_name)

        if self.transform:
            image = self.transform(image)

        label = self.labels[idx]

        return image, label

先创建Dataset，再创建dataloader，从Dataset取minibatch。

    # 假设 image_dir 是包含所有图像文件的文件夹路径，labels 是标签列表
    train_set = CustomDataset(img_dir, img_names_rgb[-500:], labels_rgb[-500:], transform=transform)
    val_set = CustomDataset(img_dir, img_names_rgb[-500:-100], labels_rgb[-500:-100], transform=transform)
    test_set = CustomDataset(img_dir, img_names_rgb[-100:], labels_rgb[-100:], transform=transform)

    # 创建一个 DataLoader
    train_loader = DataLoader(train_set, batch_size=32, shuffle=True, drop_last=False)
    val_loader = DataLoader(val_set, batch_size=32, shuffle=True, drop_last=False, num_workers=4)
    test_loader = DataLoader(test_set, batch_size=32, shuffle=True, drop_last=False, num_workers=4)

以上，数据准备工作完成，对模型进行训练

    trainer = L.Trainer(
            default_root_dir=os.path.join(CHECKPOINT_PATH, "ViT"),
            accelerator="auto",
            devices=1,
            max_epochs=180,
            callbacks=[
                ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc"),
                LearningRateMonitor("epoch"),
            ],
        )
    trainer.logger._log_graph = True  # If True, we plot the computation graph in tensorboard
    trainer.logger._default_hp_metric = None  # Optional logging argument that we don't need

    trainer.fit(model, train_loader, val_loader)
    # Load best checkpoint after training
    model = ViT.load_from_checkpoint(trainer.checkpoint_callback.best_model_path)

    # Test best model on validation and test set
    val_result = trainer.test(model, dataloaders=val_loader, verbose=False)
    test_result = trainer.test(model, dataloaders=test_loader, verbose=False)
    result = {"test": test_result[0]["test_acc"], "val": val_result[0]["test_acc"]}

完整代码：

一共包括三个文件：model.py 搭建网络的功能, prepare_data.py 数据预处理工作, main.py 网络训练

model.py内容：

import torch.nn as nn
import torch
import torch.optim as optim
import torch.nn.functional as F
import lightning as L


class AttentionBlock(nn.Module):
    def __init__(self, embed_dim, hidden_dim, num_heads, dropout=0.0):
        """
        Inputs:
            embed_dim - Dimensionality of input and attention feature vectors
            hidden_dim - Dimensionality of hidden layer in feed-forward network
                         (usually 2-4x larger than embed_dim)
            num_heads - Number of heads to use in the Multi-Head Attention block
            dropout - Amount of dropout to apply in the feed-forward network
        """
        super().__init__()

        self.layer_norm_1 = nn.LayerNorm(embed_dim)
        self.attn = nn.MultiheadAttention(embed_dim, num_heads)
        self.layer_norm_2 = nn.LayerNorm(embed_dim)
        self.linear = nn.Sequential(
            nn.Linear(embed_dim, hidden_dim),
            nn.GELU(),
            nn.Dropout(dropout),
            nn.Linear(hidden_dim, embed_dim),
            nn.Dropout(dropout),
        )

    def forward(self, x):
        inp_x = self.layer_norm_1(x)
        x = x + self.attn(inp_x, inp_x, inp_x)[0]
        x = x + self.linear(self.layer_norm_2(x))
        return x


class VisionTransformer(nn.Module):
    def __init__(
        self,
        embed_dim,
        hidden_dim,
        num_channels,
        num_heads,
        num_layers,
        num_classes,
        patch_size,
        num_patches,
        dropout=0.0,
    ):
        """
        Inputs:
            embed_dim - Dimensionality of the input feature vectors to the Transformer
            hidden_dim - Dimensionality of the hidden layer in the feed-forward networks
                         within the Transformer
            num_channels - Number of channels of the input (3 for RGB)
            num_heads - Number of heads to use in the Multi-Head Attention block
            num_layers - Number of layers to use in the Transformer
            num_classes - Number of classes to predict
            patch_size - Number of pixels that the patches have per dimension
            num_patches - Maximum number of patches an image can have
            dropout - Amount of dropout to apply in the feed-forward network and
                      on the input encoding
        """
        super().__init__()

        self.patch_size = patch_size

        # Layers/Networks
        self.input_layer = nn.Linear(num_channels * (patch_size**2), embed_dim)
        self.transformer = nn.Sequential(
            *(AttentionBlock(embed_dim, hidden_dim, num_heads, dropout=dropout) for _ in range(num_layers))
        )
        self.mlp_head = nn.Sequential(nn.LayerNorm(embed_dim), nn.Linear(embed_dim, num_classes))
        self.dropout = nn.Dropout(dropout)

        # Parameters/Embeddings
        self.cls_token = nn.Parameter(torch.randn(1, 1, embed_dim))
        self.pos_embedding = nn.Parameter(torch.randn(1, 1 + num_patches, embed_dim))

    def img_to_patch(self, x, patch_size, flatten_channels=True):
        """
        Inputs:
            x - Tensor representing the image of shape [B, C, H, W]
            patch_size - Number of pixels per dimension of the patches (integer)
            flatten_channels - If True, the patches will be returned in a flattened format
                               as a feature vector instead of a image grid.
        """
        B, C, H, W = x.shape
        x = x.reshape(B, C, H // patch_size, patch_size, W // patch_size, patch_size)
        x = x.permute(0, 2, 4, 1, 3, 5)  # [B, H', W', C, p_H, p_W]
        x = x.flatten(1, 2)  # [B, H'*W', C, p_H, p_W]
        if flatten_channels:
            x = x.flatten(2, 4)  # [B, H'*W', C*p_H*p_W]
        return x

    def forward(self, x):
        # Preprocess input
        x = self.img_to_patch(x, self.patch_size)
        B, T, _ = x.shape
        x = self.input_layer(x)

        # Add CLS token and positional encoding
        cls_token = self.cls_token.repeat(B, 1, 1)
        x = torch.cat([cls_token, x], dim=1)
        x = x + self.pos_embedding[:, : T + 1]

        # Apply Transforrmer
        x = self.dropout(x)
        x = x.transpose(0, 1)
        x = self.transformer(x)

        # Perform classification prediction
        cls = x[0]
        out = self.mlp_head(cls)
        return out


class ViT(L.LightningModule):
    def __init__(self, model_kwargs, lr):
        super().__init__()
        self.save_hyperparameters()
        self.model = VisionTransformer(**model_kwargs)

    def forward(self, x):
        return self.model(x)

    def configure_optimizers(self):
        optimizer = optim.AdamW(self.parameters(), lr=self.hparams.lr)
        lr_scheduler = optim.lr_scheduler.MultiStepLR(optimizer, milestones=[100, 150], gamma=0.1)
        return [optimizer], [lr_scheduler]

    def _calculate_loss(self, batch, mode="train"):
        imgs, labels = batch
        preds = self.model(imgs)
        loss = F.cross_entropy(preds, labels)
        acc = (preds.argmax(dim=-1) == labels).float().mean()

        self.log("%s_loss" % mode, loss)
        self.log("%s_acc" % mode, acc)
        return loss

    def training_step(self, batch, batch_idx):
        loss = self._calculate_loss(batch, mode="train")
        return loss

    def validation_step(self, batch, batch_idx):
        self._calculate_loss(batch, mode="val")

    def test_step(self, batch, batch_idx):
        self._calculate_loss(batch, mode="test")

prepare_data.py内容：

import os
import json
import torch
from PIL import Image
from torch.utils.data import Dataset, DataLoader
import torchvision.transforms as transforms


class CustomDataset(Dataset):
    def __init__(self, image_dir, names, labels, transform=None):
        self.image_dir = image_dir
        self.names = names
        self.labels = labels
        self.transform = transform

    def __len__(self):
        return len(self.labels)

    def __getitem__(self, idx):
        name_ = self.names[idx]
        img_name = os.path.join(self.image_dir, name_)
        image = Image.open(img_name)

        if self.transform:
            image = self.transform(image)

        label = self.labels[idx]

        return image, label


def load_img_ann(ann_path):
    """return [{img_name, [ (x, y, h, w, label), ... ]}]"""
    with open(ann_path) as fp:
        root = json.load(fp)
    img_dict = {}
    img_label_dict = {}
    for img_info in root['images']:
        img_id = img_info['id']
        img_name = img_info['file_name']
        img_dict[img_id] = {'name': img_name}
    for ann_info in root['annotations']:
        img_id = ann_info['image_id']
        img_category_id = ann_info['category_id']
        img_name = img_dict[img_id]['name']

        img_label_dict[img_id] = {'name': img_name, 'category_id': img_category_id}

    return img_label_dict


def get_dataloader():
    annota_dir = '/home/robotics/Downloads/coco_dataset/annotations/instances_val2017.json'
    img_dir = "/home/robotics/Downloads/coco_dataset/val2017"
    img_dict = load_img_ann(annota_dir)
    values = list(img_dict.values())
    img_names = []
    labels = []
    for item in values:
        category_id = item['category_id']
        labels.append(category_id)
        img_name = item['name']
        img_names.append(img_name)

    # 检查剔除黑白的图片
    img_names_rgb = []
    labels_rgb = []
    for i in range(len(img_names)):
        # 检查文件扩展名，确保它是图片文件（可以根据需要扩展支持的文件类型）
        file_path = os.path.join(img_dir, img_names[i])

        # 打开图片文件
        img = Image.open(file_path)

        # 获取通道数
        num_channels = img.mode
        if num_channels == "RGB" and labels[i] < 10:
            img_names_rgb.append(img_names[i])
            labels_rgb.append(labels[i])

    # 定义一系列图像转换操作
    transform = transforms.Compose([
        transforms.Resize((16, 16)),  # 调整图像大小
        transforms.RandomHorizontalFlip(),  # 随机水平翻转
        transforms.ToTensor(),  # 将图像转换为张量
        transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])  # 标准化图像
    ])

    # 假设 image_dir 是包含所有图像文件的文件夹路径，labels 是标签列表
    train_set = CustomDataset(img_dir, img_names_rgb[-500:], labels_rgb[-500:], transform=transform)
    val_set = CustomDataset(img_dir, img_names_rgb[-500:-100], labels_rgb[-500:-100], transform=transform)
    test_set = CustomDataset(img_dir, img_names_rgb[-100:], labels_rgb[-100:], transform=transform)

    # 创建一个 DataLoader
    train_loader = DataLoader(train_set, batch_size=32, shuffle=True, drop_last=False)
    val_loader = DataLoader(val_set, batch_size=32, shuffle=True, drop_last=False, num_workers=4)
    test_loader = DataLoader(test_set, batch_size=32, shuffle=True, drop_last=False, num_workers=4)

    return train_loader, val_loader, test_loader


if __name__ == "__main__":
    train_loader, val_loader, test_loader = get_dataloader()
    for batch in train_loader:
        print(batch)

main.py内容：

import os
os.environ['CUDA_LAUNCH_BLOCKING'] = '1' # 下面老是报错 shape 不一致
import urllib.request
from urllib.error import HTTPError
import lightning as L
from model import ViT
from torchview import draw_graph
from lightning.pytorch.callbacks import LearningRateMonitor, ModelCheckpoint

from prepare_data import get_dataloader


# 加载模型
# Files to download
base_url = "https://raw.githubusercontent.com/phlippe/saved_models/main/"
CHECKPOINT_PATH = os.environ.get("PATH_CHECKPOINT", "saved_models/VisionTransformers/")
pretrained_files = [
    "tutorial15/ViT.ckpt",
    "tutorial15/tensorboards/ViT/events.out.tfevents.ViT",
    "tutorial5/tensorboards/ResNet/events.out.tfevents.resnet",
]
# Create checkpoint path if it doesn't exist yet
os.makedirs(CHECKPOINT_PATH, exist_ok=True)

# For each file, check whether it already exists. If not, try downloading it.
for file_name in pretrained_files:
    file_path = os.path.join(CHECKPOINT_PATH, file_name.split("/", 1)[1])
    if "/" in file_name.split("/", 1)[1]:
        os.makedirs(file_path.rsplit("/", 1)[0], exist_ok=True)
    if not os.path.isfile(file_path):
        file_url = base_url + file_name
        print("Downloading %s..." % file_url)
        try:
            urllib.request.urlretrieve(file_url, file_path)
        except HTTPError as e:
            print(
                "Something went wrong. Please try to download the file from the GDrive folder, or contact the author with the full output including the following error:\n",
                e,
            )

pretrained_filename = os.path.join(CHECKPOINT_PATH, "ViT.ckpt")
needTrain = False
if not os.path.isfile(pretrained_filename):
    print("Found pretrained model at %s, loading..." % pretrained_filename)
    # Automatically loads the model with the saved hyperparameters
    model = ViT.load_from_checkpoint(pretrained_filename)
else:
    L.seed_everything(42)  # To be reproducable
    model = ViT(model_kwargs={
        "embed_dim": 256,
        "hidden_dim": 512,
        "num_heads": 8,
        "num_layers": 6,
        "patch_size": 4,
        "num_channels": 3,
        "num_patches": 64,
        "num_classes": 10,
        "dropout": 0.2,
        },
        lr=3e-4,
    )
    needTrain = True

# 网络结构可视化
model_graph = draw_graph(model, input_size=(1, 3, 16, 16))
model_graph.resize_graph(scale=5.0)
model_graph.visual_graph.render(format='svg')

# 准备训练数据
train_loader, val_loader, test_loader = get_dataloader()

if needTrain:
    trainer = L.Trainer(
            default_root_dir=os.path.join(CHECKPOINT_PATH, "ViT"),
            accelerator="auto",
            devices=1,
            max_epochs=180,
            callbacks=[
                ModelCheckpoint(save_weights_only=True, mode="max", monitor="val_acc"),
                LearningRateMonitor("epoch"),
            ],
        )
    trainer.logger._log_graph = True  # If True, we plot the computation graph in tensorboard
    trainer.logger._default_hp_metric = None  # Optional logging argument that we don't need

    trainer.fit(model, train_loader, val_loader)
    # Load best checkpoint after training
    model = ViT.load_from_checkpoint(trainer.checkpoint_callback.best_model_path)

    # Test best model on validation and test set
    val_result = trainer.test(model, dataloaders=val_loader, verbose=False)
    test_result = trainer.test(model, dataloaders=test_loader, verbose=False)
    result = {"test": test_result[0]["test_acc"], "val": val_result[0]["test_acc"]}