CNN PyTorch Fixture Import

CNN fixture weight import from a PyTorch-style state_dict.

We mirror the common PyTorch naming convention for modules:

• conv1.weight, conv1.bias
• conv2.weight, conv2.bias
• fc.weight, fc.bias

Each tensor is expected to be encoded as nested JSON arrays whose shape matches the expected TorchLean Shape.

structure Import.CNNPyTorch.CnnStateDict (inC outC kH kW flatSize : ℕ) : Type

Typed view of a PyTorch state_dict for the demo 2-block CNN.

This matches the keys used by the exporter (conv1.*, conv2.*, fc.*) and pins down the exact shapes expected by TorchLean.

• convW1 : Spec.Tensor Float (Spec.Shape.dim outC (Spec.Shape.dim inC (Spec.Shape.dim kH (Spec.Shape.dim kW Spec.Shape.scalar))))

  First convolution kernel, PyTorch shape (outC, inC, kH, kW).

• convB1 : Spec.Tensor Float (Spec.Shape.dim outC Spec.Shape.scalar)

  First convolution bias.

• convW2 : Spec.Tensor Float (Spec.Shape.dim outC (Spec.Shape.dim outC (Spec.Shape.dim kH (Spec.Shape.dim kW Spec.Shape.scalar))))

  Second convolution kernel, PyTorch shape (outC, outC, kH, kW).

• convB2 : Spec.Tensor Float (Spec.Shape.dim outC Spec.Shape.scalar)

  Second convolution bias.

• linearW : Spec.Tensor Float (Spec.Shape.dim outC (Spec.Shape.dim flatSize Spec.Shape.scalar))

  Classifier weight, PyTorch shape (outC, flatSize).

• linearB : Spec.Tensor Float (Spec.Shape.dim outC Spec.Shape.scalar)

  Classifier bias.

def Import.CNNPyTorch.loadCnnStateDict (inC outC kH kW flatSize : ℕ) (j : Lean.Json) : Option (CnnStateDict inC outC kH kW flatSize)

Load a CNN state dict from JSON (PyTorch state_dict-style keys).