pytorch/pytorch
Packages
A few torch.* packages are big enough to warrant their own page even though they're not "systems" in the architectural sense. Each is a substantial, user-facing surface area.
| Page | What it is |
|---|---|
torch.nn |
Neural network module library |
torch.optim |
Optimizers and learning-rate schedulers |
torch.utils.data |
DataLoader, Dataset, Sampler, multi-worker loading |
torch.func |
Function transforms (vmap, grad, jvp, vjp, …) |
torch.linalg |
Numerical linear algebra |
These are documented as packages because they're how a typical user interacts with PyTorch. The systems they sit on top of are documented in Systems.
Built by Factory AutoWiki from public repository content. It is a generated preview for codebase exploration, not source-maintained documentation.