Reference and Portable Backend Capsules

Portable capsules for paths that do not require CUDA or LibTorch.

These are not meant to win benchmarks. They keep TorchLean runnable on CPU-only machines and give the planner a clear fallback vocabulary for reference/spec-aligned execution.

def NN.Backend.Reference.referenceCapsule (name : String) (op : BackendOp) (specName valueSummary vjpSummary : String) (vjpMode : VJPMode := VJPMode.torchLeanTape) : KernelCapsule

def NN.Backend.Reference.referencePointwiseCapsule (op : BackendOp) : KernelCapsule

def NN.Backend.Reference.referenceReductionCapsule (op : BackendOp) : KernelCapsule

def NN.Backend.Reference.referenceViewCapsule (op : BackendOp) : KernelCapsule

def NN.Backend.Reference.referenceForwardOnlyCapsule (op : BackendOp) (specName valueSummary : String) : KernelCapsule

def NN.Backend.Reference.referenceConvPoolCapsule (op : BackendOp) : KernelCapsule

def NN.Backend.Reference.relu : KernelCapsule

Reference ReLU activation.

def NN.Backend.Reference.gelu : KernelCapsule

Reference GELU activation.

def NN.Backend.Reference.add : KernelCapsule

def NN.Backend.Reference.sub : KernelCapsule

def NN.Backend.Reference.mul : KernelCapsule

def NN.Backend.Reference.scale : KernelCapsule

def NN.Backend.Reference.abs : KernelCapsule

def NN.Backend.Reference.sqrt : KernelCapsule

def NN.Backend.Reference.clamp : KernelCapsule

def NN.Backend.Reference.max : KernelCapsule

def NN.Backend.Reference.min : KernelCapsule

def NN.Backend.Reference.sigmoid : KernelCapsule

def NN.Backend.Reference.tanh : KernelCapsule

def NN.Backend.Reference.softplus : KernelCapsule

def NN.Backend.Reference.exp : KernelCapsule

def NN.Backend.Reference.log : KernelCapsule

def NN.Backend.Reference.sin : KernelCapsule

def NN.Backend.Reference.cos : KernelCapsule

def NN.Backend.Reference.inv : KernelCapsule

def NN.Backend.Reference.safeLog : KernelCapsule

def NN.Backend.Reference.logSoftmax : KernelCapsule

def NN.Backend.Reference.softmax : KernelCapsule

Reference softmax path.

def NN.Backend.Reference.sum : KernelCapsule

def NN.Backend.Reference.reduceSum : KernelCapsule

def NN.Backend.Reference.reduceMean : KernelCapsule

def NN.Backend.Reference.flatten : KernelCapsule

def NN.Backend.Reference.reshape : KernelCapsule

def NN.Backend.Reference.permute : KernelCapsule

def NN.Backend.Reference.transpose2d : KernelCapsule

def NN.Backend.Reference.swapAdjacentAtDepth : KernelCapsule

def NN.Backend.Reference.transpose3dFirstToLast : KernelCapsule

def NN.Backend.Reference.transpose3dLastToFirst : KernelCapsule

def NN.Backend.Reference.transpose3dLastTwo : KernelCapsule

def NN.Backend.Reference.broadcastTo : KernelCapsule

def NN.Backend.Reference.concatVectors : KernelCapsule

def NN.Backend.Reference.concatLeadingAxis : KernelCapsule

def NN.Backend.Reference.sliceLeadingAxisRange : KernelCapsule

def NN.Backend.Reference.gatherScalar : KernelCapsule

def NN.Backend.Reference.gatherScalarNat : KernelCapsule

def NN.Backend.Reference.gatherRow : KernelCapsule

def NN.Backend.Reference.gatherVecNat : KernelCapsule

def NN.Backend.Reference.gatherRowsNat : KernelCapsule

def NN.Backend.Reference.scatterAddVec : KernelCapsule

def NN.Backend.Reference.scatterAddRow : KernelCapsule

def NN.Backend.Reference.randUniform : KernelCapsule

def NN.Backend.Reference.bernoulliMask : KernelCapsule

def NN.Backend.Reference.matmul : KernelCapsule

Reference matmul path.

def NN.Backend.Reference.bmm : KernelCapsule

Reference batched matrix multiplication.

def NN.Backend.Reference.linear : KernelCapsule

Reference linear layer.

def NN.Backend.Reference.mseLoss : KernelCapsule

Reference mean-squared-error loss.

def NN.Backend.Reference.layerNorm : KernelCapsule

Reference layer normalization.

def NN.Backend.Reference.batchNorm : KernelCapsule

Reference batch normalization.

def NN.Backend.Reference.batchNormChannelFirst : KernelCapsule

Reference channel-first BatchNorm runtime op.

def NN.Backend.Reference.conv : KernelCapsule

Reference generic channel-first convolution.

def NN.Backend.Reference.conv2d : KernelCapsule

Reference 2D convolution.

def NN.Backend.Reference.convTranspose : KernelCapsule

Reference generic channel-first transpose convolution.

def NN.Backend.Reference.convTranspose2d : KernelCapsule

Reference 2D transpose convolution.

def NN.Backend.Reference.maxPool : KernelCapsule

Reference max pooling.

def NN.Backend.Reference.maxPool2d : KernelCapsule

Reference 2D max pooling.

def NN.Backend.Reference.maxPool2dPad : KernelCapsule

Reference padded 2D max pooling.

def NN.Backend.Reference.smoothMaxPool : KernelCapsule

Reference smooth max pooling.

def NN.Backend.Reference.smoothMaxPool2d : KernelCapsule

Reference smooth 2D max pooling.

def NN.Backend.Reference.avgPool : KernelCapsule

Reference average pooling.

def NN.Backend.Reference.avgPool2d : KernelCapsule

Reference 2D average pooling.

def NN.Backend.Reference.avgPool2dPad : KernelCapsule

Reference padded 2D average pooling.

def NN.Backend.Reference.attention : KernelCapsule

Reference attention path using the composed TorchLean expression.

def NN.Backend.Reference.capsules : List KernelCapsule

Cross-platform reference capsules.