juniorPyTorch
What is a tensor in PyTorch?
Updated May 17, 2026
Short answer
A tensor is the core data structure in PyTorch, similar to NumPy arrays but with GPU support and autograd capabilities.
Deep explanation
Tensors are multi-dimensional arrays that support automatic differentiation. They are the foundation of all PyTorch computations and can be moved between CPU and GPU seamlessly.
Real-world example
Used to represent images, text embeddings, and model parameters in deep learning models.
Common mistakes
- Confusing tensors with Python lists or forgetting requires_grad for training variables.
Follow-up questions
- What is tensor broadcasting?
- How are tensors stored in memory?