Deep Learning with PyTorch
A complete 25-lesson Deep Learning course with PyTorch — from tensors, autograd, and backpropagation through CNNs, RNNs, attention, and transformers, culminating in an image classifier, a text classifier, and a mini transformer built from scratch.
25 Lessons · 750 min · Learning Platform Editorial Team
What you need on the desk
- Python 3 with PyTorch installed (CPU-only is fine for every example in this course)
- A code editor (VS Code recommended)
- No GPU or paid cloud account required — every example runs on CPU, just slower
Know this much before you start
- Comfort with Python and basic NumPy/pandas (the Python, NumPy, and pandas courses on this site cover this)
- Comfort with the Machine Learning Basics course's ideas (train/test split, overfitting) is helpful but not required — this course re-explains what it needs
- High-school-level math comfort with functions and derivatives; the calculus itself is explained as it's used, not assumed
- No prior deep learning or PyTorch experience needed
By the end, you can
- Explain tensors, autograd, and backpropagation, and implement a training loop from scratch
- Build and train feedforward, convolutional, and recurrent neural networks in PyTorch
- Diagnose overfitting and apply regularization, normalization, and learning-rate scheduling correctly
- Explain the attention mechanism and the transformer architecture that power modern LLMs
- Evaluate models with the right metrics, and save/load/deploy a trained model for inference
- Build and test real small deep learning projects — an image classifier, a text classifier, and a mini transformer