ndarray
NumPy's N-dimensional array - homogeneous, fixed-size, contiguous memory.
- Bundles memory + shape + dtype + strides.
- Single dtype enables fast C dispatch.
- View vs. copy semantics drive performance.
When to use: Any numerical pipeline; preprocessing, batches, intermediate tensors.
Example: np.zeros((32, 3, 224, 224), dtype=np.float32) → 32-image batch, ~19 MB.