Runtime

Import this file when you need TorchLean's executable layer. It collects the runtime pieces used for building, training, importing, exporting, or checking runnable models:

• the eager and compiled autograd engines;
• the lower-level Runtime.Autograd.Torch session operations;
• the higher-level Runtime.Autograd.TorchLean front-end used by NN.API.Runtime;
• deterministic dataset/training utilities;
• optional external-process helpers for untrusted producer / trusted checker workflows;
• pure optimizer and scheduler equations;
• PyTorch import/export bridge infrastructure; and
• typed reinforcement-learning runtime helpers.

For ordinary application code, prefer import NN.API. Import this file when you need the full executable subsystem. If you only need pure tensor semantics and theorems, prefer NN.Spec or NN.Proofs; those imports keep runtime bridge dependencies out of the build.

The runtime import includes reusable bridge infrastructure under NN.Runtime.PyTorch.*. Example-only MLP/CNN/Transformer round-trip code lives under NN.Examples.Interop.PyTorch.*, so ordinary runtime imports do not pull example modules into the library API.

References / context:

• PyTorch autograd overview: https://pytorch.org/docs/stable/autograd.html
• PyTorch nn.Module / tensor ops API: https://pytorch.org/docs/stable/nn.html https://pytorch.org/docs/stable/torch.html
• TorchLean’s import/export bridge details live in NN.Runtime.PyTorch.Export.Core and NN.Runtime.PyTorch.Import.Core.