train.py

# train.py
import torch
import pandas as pd
import numpy as np
from torch import nn
from torch.utils.data import Dataset, DataLoader
from sklearn.model_selection import train_test_split

class MangaDataset(Dataset):
    def __init__(self, ratings_df):
        self.users = torch.tensor(ratings_df['user_idx'].values, dtype=torch.long)
        self.items = torch.tensor(ratings_df['manga_idx'].values, dtype=torch.long)
        self.ratings = torch.tensor(ratings_df['rating'].values, dtype=torch.float)
        
    def __len__(self):
        return len(self.ratings)
    
    def __getitem__(self, idx):
        return self.users[idx], self.items[idx], self.ratings[idx]

class MangaRecommender(nn.Module):
    def __init__(self, num_users, num_items, n_factors=50):
        super().__init__()
        self.user_factors = nn.Embedding(num_users, n_factors)
        self.item_factors = nn.Embedding(num_items, n_factors)
        
        # Initialize embeddings
        nn.init.xavier_normal_(self.user_factors.weight)
        nn.init.xavier_normal_(self.item_factors.weight)
        
    def forward(self, user, item):
        user_emb = self.user_factors(user)
        item_emb = self.item_factors(item)
        return (user_emb * item_emb).sum(1)
    
    def predict(self, user_ids):
        user_emb = self.user_factors(user_ids)
        all_items = self.item_factors.weight
        return torch.matmul(user_emb, all_items.t())

def train_model():
    # Load your data
    df = pd.read_csv('manga_ratings.csv')
    
    # Create user and item mappings
    user_mapping = {uid: idx for idx, uid in enumerate(df['user_id'].unique())}
    manga_mapping = {mid: idx for idx, mid in enumerate(df['manga_id'].unique())}
    
    # Convert ratings to numerical values
    rating_map = {'like': 1.0, 'dislike': -1.0, None: 0.0}
    
    # Prepare training data
    df['user_idx'] = df['user_id'].map(user_mapping)
    df['manga_idx'] = df['manga_id'].map(manga_mapping)
    df['rating'] = df['like_status'].map(rating_map)
    
    # Create train/val split
    train_df, val_df = train_test_split(df, test_size=0.2)
    
    # Create datasets
    train_dataset = MangaDataset(train_df)
    val_dataset = MangaDataset(val_df)
    
    # Create dataloaders
    train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)
    val_loader = DataLoader(val_dataset, batch_size=64)
    
    # Initialize model
    model = MangaRecommender(
        num_users=len(user_mapping),
        num_items=len(manga_mapping)
    )
    
    # Training setup
    criterion = nn.MSELoss()
    optimizer = torch.optim.Adam(model.parameters(), lr=0.001)
    
    # Training loop
    num_epochs = 20
    for epoch in range(num_epochs):
        model.train()
        total_loss = 0
        for user, item, rating in train_loader:
            optimizer.zero_grad()
            pred = model(user, item)
            loss = criterion(pred, rating)
            loss.backward()
            optimizer.step()
            total_loss += loss.item()
            
        # Validation
        model.eval()
        val_loss = 0
        with torch.no_grad():
            for user, item, rating in val_loader:
                pred = model(user, item)
                val_loss += criterion(pred, rating).item()
                
        print(f'Epoch {epoch+1}/{num_epochs}')
        print(f'Train Loss: {total_loss/len(train_loader):.4f}')
        print(f'Val Loss: {val_loss/len(val_loader):.4f}')
    
    # Save mappings and model
    torch.save({
        'model_state_dict': model.state_dict(),
        'user_mapping': user_mapping,
        'manga_mapping': manga_mapping
    }, 'manga_recommender.pt')

if __name__ == '__main__':
    train_model()