【Pytorch】模块化

---

模块化涉及将jupyter notebook代码转换为一系列提供类似功能的不同 Python 脚本。

可以将笔记本代码从一系列单元格转换为以下 Python 文件：

1. data_setup.py - 如果需要，用于准备和下载数据的文件。
2. engine.py - 包含各种训练函数的文件。
3. model_builder.py 或 model.py - 用于创建 PyTorch 模型的文件。
4. train.py - 用于利用所有其他文件并训练目标 PyTorch 模型的文件。
5. utils.py - 专用于有用实用功能的文件。

上述文件的命名和布局将取决于您的用例和代码要求。 Python 脚本与单个notebook单元一样通用，这意味着您可以为几乎任何类型的功能创建脚本。

notebook非常适合快速迭代探索和运行实验，但是，对于较大规模的项目，您可能会发现 Python 脚本更具可重复性且更易于运行。

在你去download别人开源的项目时，可能会指示您在终端/命令行中运行如下代码来训练模型：

python train.py --model MODEL_NAME --batch_size BATCH_SIZE --lr LEARNING_RATE --num_epochs NUM_EPOCHS

train.py 是目标 Python 脚本，它可能包含训练 PyTorch 模型的函数，--model 、 --batch_size 、 --lr 和 --num_epochs 被称为参数标志。

可以将它们设置为您喜欢的任何值，如果它们与 train.py 兼容，它们就会工作，如果不兼容，它们就会出错。

例如，训练 TinyVGG 模型 10 个时期，批量大小为 32，学习率为 0.001：

python train.py --model tinyvgg --batch_size 32 --lr 0.001 --num_epochs 10

Python脚本的目录结构：

going_modular/
├── going_modular/
│   ├── data_setup.py
│   ├── engine.py
│   ├── model_builder.py
│   ├── train.py
│   └── utils.py
├── models/
│   ├── 05_going_modular_cell_mode_tinyvgg_model.pth
│   └── 05_going_modular_script_mode_tinyvgg_model.pth
└── data/
    └── pizza_steak_sushi/
        ├── train/
        │   ├── pizza/
        │   │   ├── image01.jpeg
        │   │   └── ...
        │   ├── steak/
        │   └── sushi/
        └── test/
            ├── pizza/
            ├── steak/
            └── sushi/

---

1. 获取数据

---

2. 创建Dataset和DataLoader

( data_setup.py )

"""
Contains functionality for creating PyTorch DataLoaders for 
image classification data.
"""
import os

from torchvision import datasets, transforms
from torch.utils.data import DataLoader

NUM_WORKERS = os.cpu_count()

def create_dataloaders(
    train_dir: str, 
    test_dir: str, 
    transform: transforms.Compose, 
    batch_size: int, 
    num_workers: int=NUM_WORKERS
):
  """Creates training and testing DataLoaders.

  Takes in a training directory and testing directory path and turns
  them into PyTorch Datasets and then into PyTorch DataLoaders.

  Args:
    train_dir: Path to training directory.
    test_dir: Path to testing directory.
    transform: torchvision transforms to perform on training and testing data.
    batch_size: Number of samples per batch in each of the DataLoaders.
    num_workers: An integer for number of workers per DataLoader.

  Returns:
    A tuple of (train_dataloader, test_dataloader, class_names).
    Where class_names is a list of the target classes.
    Example usage:
      train_dataloader, test_dataloader, class_names = \
        = create_dataloaders(train_dir=path/to/train_dir,
                             test_dir=path/to/test_dir,
                             transform=some_transform,
                             batch_size=32,
                             num_workers=4)
  """
  # Use ImageFolder to create dataset(s)
  train_data = datasets.ImageFolder(train_dir, transform=transform)
  test_data = datasets.ImageFolder(test_dir, transform=transform)

  # Get class names
  class_names = train_data.classes

  # Turn images into data loaders
  train_dataloader = DataLoader(
      train_data,
      batch_size=batch_size,
      shuffle=True,
      num_workers=num_workers,
      pin_memory=True,
  )
  test_dataloader = DataLoader(
      test_data,
      batch_size=batch_size,
      shuffle=False,
      num_workers=num_workers,
      pin_memory=True,
  )

  return train_dataloader, test_dataloader, class_names

如果我们想要创建 DataLoader ，我们现在可以在 data_setup.py 中使用该函数，如下所示：

# Import data_setup.py
from going_modular import data_setup

# Create train/test dataloader and get class names as a list
train_dataloader, test_dataloader, class_names = data_setup.create_dataloaders(...)

---

3. 定义模型

（model_builder.py）

"""
Contains PyTorch model code to instantiate a TinyVGG model.
"""
import torch
from torch import nn 

class TinyVGG(nn.Module):
    """Creates the TinyVGG architecture.

    Replicates the TinyVGG architecture from the CNN explainer website in PyTorch.
    See the original architecture here: https://poloclub.github.io/cnn-explainer/

    Args:
    input_shape: An integer indicating number of input channels.
    hidden_units: An integer indicating number of hidden units between layers.
    output_shape: An integer indicating number of output units.
    """
    def __init__(self, input_shape: int, hidden_units: int, output_shape: int) -> None:
        super().__init__()
        self.conv_block_1 = nn.Sequential(
          nn.Conv2d(in_channels=input_shape, 
                    out_channels=hidden_units, 
                    kernel_size=3, 
                    stride=1, 
                    padding=0),  
          nn.ReLU(),
          nn.Conv2d(in_channels=hidden_units, 
                    out_channels=hidden_units,
                    kernel_size=3,
                    stride=1,
                    padding=0),
          nn.ReLU(),
          nn.MaxPool2d(kernel_size=2,
                        stride=2)
        )
        self.conv_block_2 = nn.Sequential(
          nn.Conv2d(hidden_units, hidden_units, kernel_size=3, padding=0),
          nn.ReLU(),
          nn.Conv2d(hidden_units, hidden_units, kernel_size=3, padding=0),
          nn.ReLU(),
          nn.MaxPool2d(2)
        )
        self.classifier = nn.Sequential(
          nn.Flatten(),
          # Where did this in_features shape come from? 
          # It's because each layer of our network compresses and changes the shape of our inputs data.
          nn.Linear(in_features=hidden_units*13*13,
                    out_features=output_shape)
        )
    
    def forward(self, x: torch.Tensor):
        x = self.conv_block_1(x)
        x = self.conv_block_2(x)
        x = self.classifier(x)
        return x
        # return self.classifier(self.block_2(self.block_1(x))) # <- leverage the benefits of operator fusion

---

4. 创建训练模型引擎函数

1. train_step() - 接受模型、 DataLoader 、损失函数和优化器，并在 DataLoader 上训练模型。
2. test_step() - 接受模型、 DataLoader 和损失函数，并在 DataLoader 上评估模型。
3. train() - 对给定数量的 epoch 一起执行 1. 和 2. 并返回结果字典。

由于这些将成为我们模型训练的引擎，因此我们可以将它们全部放入名为 engine.py 的 Python 脚本中：

"""
Contains functions for training and testing a PyTorch model.
"""
import torch

from tqdm.auto import tqdm
from typing import Dict, List, Tuple

def train_step(model: torch.nn.Module, 
               dataloader: torch.utils.data.DataLoader, 
               loss_fn: torch.nn.Module, 
               optimizer: torch.optim.Optimizer,
               device: torch.device) -> Tuple[float, float]:
  """Trains a PyTorch model for a single epoch.

  Turns a target PyTorch model to training mode and then
  runs through all of the required training steps (forward
  pass, loss calculation, optimizer step).

  Args:
    model: A PyTorch model to be trained.
    dataloader: A DataLoader instance for the model to be trained on.
    loss_fn: A PyTorch loss function to minimize.
    optimizer: A PyTorch optimizer to help minimize the loss function.
    device: A target device to compute on (e.g. "cuda" or "cpu").

  Returns:
    A tuple of training loss and training accuracy metrics.
    In the form (train_loss, train_accuracy). For example:

    (0.1112, 0.8743)
  """
  # Put model in train mode
  model.train()

  # Setup train loss and train accuracy values
  train_loss, train_acc = 0, 0

  # Loop through data loader data batches
  for batch, (X, y) in enumerate(dataloader):
      # Send data to target device
      X, y = X.to(device), y.to(device)

      # 1. Forward pass
      y_pred = model(X)

      # 2. Calculate  and accumulate loss
      loss = loss_fn(y_pred, y)
      train_loss += loss.item() 

      # 3. Optimizer zero grad
      optimizer.zero_grad()

      # 4. Loss backward
      loss.backward()

      # 5. Optimizer step
      optimizer.step()

      # Calculate and accumulate accuracy metric across all batches
      y_pred_class = torch.argmax(torch.softmax(y_pred, dim=1), dim=1)
      train_acc += (y_pred_class == y).sum().item()/len(y_pred)

  # Adjust metrics to get average loss and accuracy per batch 
  train_loss = train_loss / len(dataloader)
  train_acc = train_acc / len(dataloader)
  return train_loss, train_acc

def test_step(model: torch.nn.Module, 
              dataloader: torch.utils.data.DataLoader, 
              loss_fn: torch.nn.Module,
              device: torch.device) -> Tuple[float, float]:
  """Tests a PyTorch model for a single epoch.

  Turns a target PyTorch model to "eval" mode and then performs
  a forward pass on a testing dataset.

  Args:
    model: A PyTorch model to be tested.
    dataloader: A DataLoader instance for the model to be tested on.
    loss_fn: A PyTorch loss function to calculate loss on the test data.
    device: A target device to compute on (e.g. "cuda" or "cpu").

  Returns:
    A tuple of testing loss and testing accuracy metrics.
    In the form (test_loss, test_accuracy). For example:

    (0.0223, 0.8985)
  """
  # Put model in eval mode
  model.eval() 

  # Setup test loss and test accuracy values
  test_loss, test_acc = 0, 0

  # Turn on inference context manager
  with torch.inference_mode():
      # Loop through DataLoader batches
      for batch, (X, y) in enumerate(dataloader):
          # Send data to target device
          X, y = X.to(device), y.to(device)

          # 1. Forward pass
          test_pred_logits = model(X)

          # 2. Calculate and accumulate loss
          loss = loss_fn(test_pred_logits, y)
          test_loss += loss.item()

          # Calculate and accumulate accuracy
          test_pred_labels = test_pred_logits.argmax(dim=1)
          test_acc += ((test_pred_labels == y).sum().item()/len(test_pred_labels))

  # Adjust metrics to get average loss and accuracy per batch 
  test_loss = test_loss / len(dataloader)
  test_acc = test_acc / len(dataloader)
  return test_loss, test_acc

def train(model: torch.nn.Module, 
          train_dataloader: torch.utils.data.DataLoader, 
          test_dataloader: torch.utils.data.DataLoader, 
          optimizer: torch.optim.Optimizer,
          loss_fn: torch.nn.Module,
          epochs: int,
          device: torch.device) -> Dict[str, List]:
  """Trains and tests a PyTorch model.

  Passes a target PyTorch models through train_step() and test_step()
  functions for a number of epochs, training and testing the model
  in the same epoch loop.

  Calculates, prints and stores evaluation metrics throughout.

  Args:
    model: A PyTorch model to be trained and tested.
    train_dataloader: A DataLoader instance for the model to be trained on.
    test_dataloader: A DataLoader instance for the model to be tested on.
    optimizer: A PyTorch optimizer to help minimize the loss function.
    loss_fn: A PyTorch loss function to calculate loss on both datasets.
    epochs: An integer indicating how many epochs to train for.
    device: A target device to compute on (e.g. "cuda" or "cpu").

  Returns:
    A dictionary of training and testing loss as well as training and
    testing accuracy metrics. Each metric has a value in a list for 
    each epoch.
    In the form: {train_loss: [...],
                  train_acc: [...],
                  test_loss: [...],
                  test_acc: [...]} 
    For example if training for epochs=2: 
                 {train_loss: [2.0616, 1.0537],
                  train_acc: [0.3945, 0.3945],
                  test_loss: [1.2641, 1.5706],
                  test_acc: [0.3400, 0.2973]} 
  """
  # Create empty results dictionary
  results = {"train_loss": [],
      "train_acc": [],
      "test_loss": [],
      "test_acc": []
  }

  # Loop through training and testing steps for a number of epochs
  for epoch in tqdm(range(epochs)):
      train_loss, train_acc = train_step(model=model,
                                          dataloader=train_dataloader,
                                          loss_fn=loss_fn,
                                          optimizer=optimizer,
                                          device=device)
      test_loss, test_acc = test_step(model=model,
          dataloader=test_dataloader,
          loss_fn=loss_fn,
          device=device)

      # Print out what's happening
      print(
          f"Epoch: {epoch+1} | "
          f"train_loss: {train_loss:.4f} | "
          f"train_acc: {train_acc:.4f} | "
          f"test_loss: {test_loss:.4f} | "
          f"test_acc: {test_acc:.4f}"
      )

      # Update results dictionary
      results["train_loss"].append(train_loss)
      results["train_acc"].append(train_acc)
      results["test_loss"].append(test_loss)
      results["test_acc"].append(test_acc)

  # Return the filled results at the end of the epochs
  return results

现在我们已经有了 engine.py 脚本，我们可以通过以下方式从中导入函数：

# Import engine.py
from going_modular import engine

# Use train() by calling it from engine.py
engine.train(...)

---

5. 创建保存模型的函数

( utils.py )

将 save_model() 函数保存到名为 utils.py 的文件中：

"""
Contains various utility functions for PyTorch model training and saving.
"""
import torch
from pathlib import Path

def save_model(model: torch.nn.Module,
               target_dir: str,
               model_name: str):
  """Saves a PyTorch model to a target directory.

  Args:
    model: A target PyTorch model to save.
    target_dir: A directory for saving the model to.
    model_name: A filename for the saved model. Should include
      either ".pth" or ".pt" as the file extension.

  Example usage:
    save_model(model=model_0,
               target_dir="models",
               model_name="05_going_modular_tingvgg_model.pth")
  """
  # Create target directory
  target_dir_path = Path(target_dir)
  target_dir_path.mkdir(parents=True,
                        exist_ok=True)

  # Create model save path
  assert model_name.endswith(".pth") or model_name.endswith(".pt"), "model_name should end with '.pt' or '.pth'"
  model_save_path = target_dir_path / model_name

  # Save the model state_dict()
  print(f"[INFO] Saving model to: {model_save_path}")
  torch.save(obj=model.state_dict(),
             f=model_save_path)

可以导入它并通过以下方式使用它，而不是重新编写它：

# Import utils.py
from going_modular import utils

# Save a model to file
save_model(model=...
           target_dir=...,
           model_name=...)

---

6. 训练、评估并保存模型

( train.py )
可以在命令行上使用一行代码来训练 PyTorch 模型：

python train.py

要创建 train.py ，我们将执行以下步骤：

1. 导入各种依赖项，即 torch 、 os 、 torchvision.transforms 以及 going_modular 目录 data_setup 、 model_builder 、 utils 。
2. 注意：由于 train.py 将位于 going_modular 目录中，因此我们可以通过 import … 而不是 from going_modular import … 导入其他模块。
3. 设置各种超参数，例如批量大小、时期数、学习率和隐藏单元数（将来可以通过 Python 的 argparse 设置）。
4. 设置训练和测试目录。
5. 设置与设备无关的代码。
6. 创建必要的数据转换。
7. 使用 data_setup.py 创建 DataLoaders。
8. 使用 model_builder.py 创建模型。
9. 设置损失函数和优化器。
10. 使用 engine.py 训练模型。
11. 使用 utils.py 保存模型。

"""
Trains a PyTorch image classification model using device-agnostic code.
"""

import os
import torch
import data_setup, engine, model_builder, utils

from torchvision import transforms

# Setup hyperparameters
NUM_EPOCHS = 5
BATCH_SIZE = 32
HIDDEN_UNITS = 10
LEARNING_RATE = 0.001

# Setup directories
train_dir = "data/pizza_steak_sushi/train"
test_dir = "data/pizza_steak_sushi/test"

# Setup target device
device = "cuda" if torch.cuda.is_available() else "cpu"

# Create transforms
data_transform = transforms.Compose([
  transforms.Resize((64, 64)),
  transforms.ToTensor()
])

# Create DataLoaders with help from data_setup.py
train_dataloader, test_dataloader, class_names = data_setup.create_dataloaders(
    train_dir=train_dir,
    test_dir=test_dir,
    transform=data_transform,
    batch_size=BATCH_SIZE
)

# Create model with help from model_builder.py
model = model_builder.TinyVGG(
    input_shape=3,
    hidden_units=HIDDEN_UNITS,
    output_shape=len(class_names)
).to(device)

# Set loss and optimizer
loss_fn = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(),
                             lr=LEARNING_RATE)

# Start training with help from engine.py
engine.train(model=model,
             train_dataloader=train_dataloader,
             test_dataloader=test_dataloader,
             loss_fn=loss_fn,
             optimizer=optimizer,
             epochs=NUM_EPOCHS,
             device=device)

# Save the model with help from utils.py
utils.save_model(model=model,
                 target_dir="models",
                 model_name="05_going_modular_script_mode_tinyvgg_model.pth")

可以调整 train.py 文件以使用 Python 的 argparse 模块的参数标志输入，这将允许我们提供不同的超参数设置，如前面讨论的：

python train.py --model MODEL_NAME --batch_size BATCH_SIZE --lr LEARNING_RATE --num_epochs NUM_EPOCHS

---