train_model.py

import torch
from torchvision import datasets, transforms
from torch.utils.data import DataLoader
import torch.nn as nn
import torch.optim as optim
from torchvision.models import convnext_tiny
from tqdm import tqdm
import matplotlib.pyplot as plt
import os

# Dataset path on RunPod
dataset_path = "/workspace/VCR Cleaned"
train_dir = os.path.join(dataset_path, "train")
val_dir = os.path.join(dataset_path, "val")

# Transforms
train_transform = transforms.Compose([
    transforms.Resize((512, 512)),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406],
                         [0.229, 0.224, 0.225])
])

val_transform = transforms.Compose([
    transforms.Resize((512, 512)),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406],
                         [0.229, 0.224, 0.225])
])

# Load datasets
train_dataset = datasets.ImageFolder(train_dir, transform=train_transform)
val_dataset = datasets.ImageFolder(val_dir, transform=val_transform)

# Verify class mapping
print("\nLabel mapping:", train_dataset.class_to_idx)
print("Number of classes:", len(train_dataset.classes))

# Load model
model = convnext_tiny(pretrained=True)
model.classifier[2] = nn.Linear(768, len(train_dataset.classes))

# Setup
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.AdamW(model.parameters(), lr=1e-4, weight_decay=0.01)
scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=10)

# Paths
checkpoint_path = "/workspace/convnext_checkpoint.pth"
best_model_path = "/workspace/convnext_best_model.pth"
final_model_path = "/workspace/convnext_final_model.pth"

# Load checkpoint if available
start_epoch = 0
train_losses = []
val_losses = []
val_accuracies = []
best_acc = 0.0

if os.path.exists(checkpoint_path):
    print("\nLoading checkpoint...")
    checkpoint = torch.load(checkpoint_path, map_location=device)
    model.load_state_dict(checkpoint['model_state_dict'])
    optimizer.load_state_dict(checkpoint['optimizer_state_dict'])
    scheduler.load_state_dict(checkpoint['scheduler_state_dict'])
    train_losses = checkpoint['train_losses']
    val_losses = checkpoint['val_losses']
    val_accuracies = checkpoint['val_accuracies']
    best_acc = max(val_accuracies) if val_accuracies else 0.0
    start_epoch = checkpoint['epoch']
    print(f"Resumed from epoch {start_epoch}")
else:
    print("\nStarting training from scratch")

# Training loop
for epoch in range(start_epoch, 100):
    model.train()
    train_loss = 0

    for images, labels in tqdm(DataLoader(train_dataset, batch_size=64, shuffle=True), desc=f"Epoch {epoch+1}"):
        images, labels = images.to(device), labels.to(device)

        optimizer.zero_grad()
        outputs = model(images)
        loss = criterion(outputs, labels)
        loss.backward()
        optimizer.step()
        train_loss += loss.item()

    scheduler.step()

    # Validation
    model.eval()
    val_loss = 0
    correct = 0
    with torch.no_grad():
        for images, labels in DataLoader(val_dataset, batch_size=64):
            images, labels = images.to(device), labels.to(device)
            outputs = model(images)
            val_loss += criterion(outputs, labels).item()
            preds = outputs.argmax(dim=1)
            correct += (preds == labels).sum().item()

    # Metrics
    epoch_train_loss = train_loss / len(train_dataset)
    epoch_val_loss = val_loss / len(val_dataset)
    epoch_val_acc = correct / len(val_dataset)

    train_losses.append(epoch_train_loss)
    val_losses.append(epoch_val_loss)
    val_accuracies.append(epoch_val_acc)

    print(f"\nEpoch {epoch+1}:")
    print(f"  Train Loss: {epoch_train_loss:.4f}")
    print(f"  Val Loss: {epoch_val_loss:.4f}")
    print(f"  Val Acc: {epoch_val_acc:.4f}")

    # Save checkpoint
    torch.save({
        'epoch': epoch + 1,
        'model_state_dict': model.state_dict(),
        'optimizer_state_dict': optimizer.state_dict(),
        'scheduler_state_dict': scheduler.state_dict(),
        'train_losses': train_losses,
        'val_losses': val_losses,
        'val_accuracies': val_accuracies
    }, checkpoint_path)

    # ✅ Save best model
    if epoch_val_acc > best_acc:
        best_acc = epoch_val_acc
        torch.save(model.state_dict(), best_model_path)
        print(f"✅ Best model saved at epoch {epoch+1} with acc {best_acc:.4f}")

# ✅ Save final model
torch.save(model.state_dict(), final_model_path)
print(f"\n✅ Final model saved to {final_model_path}")

# Plot
plt.figure(figsize=(12, 4))
plt.subplot(1, 2, 1)
plt.plot(train_losses, label='Train Loss')
plt.plot(val_losses, label='Val Loss')
plt.title("Loss")
plt.legend()

plt.subplot(1, 2, 2)
plt.plot(val_accuracies, label='Val Accuracy')
plt.title("Validation Accuracy")
plt.legend()
plt.show()