spdraptor/Netflix_Content_Recommender

1 2#ran on Kaggle3!pip install sentence-transformers4!pip install torch5import torch6from sentence_transformers import SentenceTransformer7import numpy as np8import pandas as pd9from tqdm import tqdm  # For tracking progress in batches10 11# Check if GPU is available12device = "cuda" if torch.cuda.is_available() else "cpu"13print(f"Using device: {device}")14 15# Load dataset16dataset = pd.read_csv('/kaggle/input/d/infamouscoder/dataset-netflix-shows/netflix_titles.csv')17 18# Load model to GPU if available19model = SentenceTransformer("all-MiniLM-L6-v2").to(device)20 21# Combine fields for embeddings22def combine_description_title_and_genre(description, listed_in, title):23    return f"{description} Genre: {listed_in} Title: {title}"24 25# Create combined text column26dataset['combined_text'] = dataset.apply(lambda row: combine_description_title_and_genre(row['description'], row['listed_in'], row['title']), axis=1)27 28# Generate embeddings in batches to save memory29batch_size = 3230embeddings = []31 32for i in tqdm(range(0, len(dataset), batch_size), desc="Generating Embeddings"):33    batch_texts = dataset['combined_text'][i:i+batch_size].tolist()34    batch_embeddings = model.encode(batch_texts, convert_to_tensor=True, device=device)35    embeddings.extend(batch_embeddings.cpu().numpy())  # Move to CPU to save memory36 37# Convert list to numpy array38embeddings = np.array(embeddings)39 40# Save embeddings and metadata41np.save("/kaggle/working/netflix_embeddings.npy", embeddings)42dataset[['show_id', 'title', 'description', 'listed_in']].to_csv("/kaggle/working/netflix_metadata.csv", index=False)