NN.Runtime.PyTorch

TorchLean’s reusable PyTorch interoperability layer.

This umbrella intentionally contains only bridge infrastructure:

• export TorchLean IR / parameters to readable PyTorch source;
• convert PyTorch state_dict checkpoints to Lean-readable JSON through a generated Python adapter; and
• capture supported PyTorch nn.Module graphs into TorchLean IR JSON; and
• parse those JSON artifacts into shape-checked TorchLean tensors, IR graphs, or verification parameter stores.

Model demos and tutorial round-trips live under NN.Examples.Interop.PyTorch.*; keeping them there prevents runtime imports from quietly depending on example-only shapes.