Deep learning frameworks provide tools to build, train, and evaluate deep learning models. They offer pre-built operations and structures to streamline the development process. ### **Why Use Deep Learning Frameworks?** - **Faster development and prototyping** with pre-built components. - **Scalable for large datasets** and complex models. - **Optimized performance** through efficient backends (e.g., CUDA, cuDNN). ### **Key Features of Deep Learning Frameworks** - **GPU Acceleration** – Speed up training with parallel computing. - **Pre-built Layers** – Simplify model construction (e.g., dense layers, convolutional layers). - **Distributed Training** – Scale training across multiple machines. - **Visualization Tools** – Helps track training progress (e.g., TensorBoard, Visdom). - **Ecosystem Integration** – Integration with libraries for NLP, reinforcement learning, etc. ### **Popular Deep Learning Frameworks** 1️. **TensorFlow** - **Created by Google** for high-performance machine learning. - Supports **neural networks, deployment, and production-level ML**. - **Keras** is a high-level API for fast model prototyping within TensorFlow. - Used in: **image recognition, NLP, Time-series forecasting**. 2️. **PyTorch** - **Developed by Facebook** and focuses on **dynamic computation graphs** (define-by-run). - Supports **GPU acceleration** and easy debugging with Python’s native tools. - Used in: **Computer vision, NLP, reinforcement learning**. ![[Frameworks.png | 500]]