Native CUDA Backend Capsules

Capsule metadata for TorchLean's native CUDA runtime provider.

These capsules describe kernels that currently live under csrc/cuda/** and are exposed to Lean through NN.Runtime.Autograd.Engine.Cuda.*. The C/CUDA source still owns the implementation; this module gives the planner a typed, inspectable contract layer for those implementation choices.

def NN.Backend.NativeCUDA.nativeCapsule (name : String) (op : BackendOp) (specName valueSummary vjpSummary : String) (vjpMode : VJPMode := VJPMode.backendVJP) : KernelCapsule

def NN.Backend.NativeCUDA.nativePointwiseCapsule (op : BackendOp) : KernelCapsule

def NN.Backend.NativeCUDA.nativeReductionCapsule (op : BackendOp) : KernelCapsule

def NN.Backend.NativeCUDA.nativeViewCapsule (op : BackendOp) : KernelCapsule

def NN.Backend.NativeCUDA.nativeForwardOnlyCapsule (op : BackendOp) (specName valueSummary : String) : KernelCapsule

def NN.Backend.NativeCUDA.nativeConvPoolCapsule (op : BackendOp) : KernelCapsule

def NN.Backend.NativeCUDA.matmul : KernelCapsule

Native CUDA batched/matrix multiplication, backed by CUDA/cuBLAS paths.

def NN.Backend.NativeCUDA.bmm : KernelCapsule

Native CUDA batched matrix multiplication.

def NN.Backend.NativeCUDA.relu : KernelCapsule

Native CUDA ReLU activation.

def NN.Backend.NativeCUDA.gelu : KernelCapsule

Native CUDA GELU activation.

def NN.Backend.NativeCUDA.add : KernelCapsule

def NN.Backend.NativeCUDA.sub : KernelCapsule

def NN.Backend.NativeCUDA.mul : KernelCapsule

def NN.Backend.NativeCUDA.scale : KernelCapsule

def NN.Backend.NativeCUDA.abs : KernelCapsule

def NN.Backend.NativeCUDA.sqrt : KernelCapsule

def NN.Backend.NativeCUDA.clamp : KernelCapsule

def NN.Backend.NativeCUDA.max : KernelCapsule

def NN.Backend.NativeCUDA.min : KernelCapsule

def NN.Backend.NativeCUDA.sigmoid : KernelCapsule

def NN.Backend.NativeCUDA.tanh : KernelCapsule

def NN.Backend.NativeCUDA.softplus : KernelCapsule

def NN.Backend.NativeCUDA.exp : KernelCapsule

def NN.Backend.NativeCUDA.log : KernelCapsule

def NN.Backend.NativeCUDA.inv : KernelCapsule

def NN.Backend.NativeCUDA.safeLog : KernelCapsule

def NN.Backend.NativeCUDA.logSoftmax : KernelCapsule

def NN.Backend.NativeCUDA.softmax : KernelCapsule

Native CUDA row/axis softmax kernels.

def NN.Backend.NativeCUDA.sum : KernelCapsule

def NN.Backend.NativeCUDA.reduceSum : KernelCapsule

def NN.Backend.NativeCUDA.reduceMean : KernelCapsule

def NN.Backend.NativeCUDA.flatten : KernelCapsule

def NN.Backend.NativeCUDA.reshape : KernelCapsule

def NN.Backend.NativeCUDA.permute : KernelCapsule

def NN.Backend.NativeCUDA.transpose2d : KernelCapsule

def NN.Backend.NativeCUDA.swapAdjacentAtDepth : KernelCapsule

def NN.Backend.NativeCUDA.transpose3dFirstToLast : KernelCapsule

def NN.Backend.NativeCUDA.transpose3dLastToFirst : KernelCapsule

def NN.Backend.NativeCUDA.transpose3dLastTwo : KernelCapsule

def NN.Backend.NativeCUDA.broadcastTo : KernelCapsule

def NN.Backend.NativeCUDA.concatVectors : KernelCapsule

def NN.Backend.NativeCUDA.concatLeadingAxis : KernelCapsule

def NN.Backend.NativeCUDA.sliceLeadingAxisRange : KernelCapsule

def NN.Backend.NativeCUDA.gatherScalar : KernelCapsule

def NN.Backend.NativeCUDA.gatherScalarNat : KernelCapsule

def NN.Backend.NativeCUDA.gatherRow : KernelCapsule

def NN.Backend.NativeCUDA.gatherVecNat : KernelCapsule

def NN.Backend.NativeCUDA.gatherRowsNat : KernelCapsule

def NN.Backend.NativeCUDA.scatterAddVec : KernelCapsule

def NN.Backend.NativeCUDA.scatterAddRow : KernelCapsule

def NN.Backend.NativeCUDA.randUniform : KernelCapsule

def NN.Backend.NativeCUDA.bernoulliMask : KernelCapsule

def NN.Backend.NativeCUDA.layerNorm : KernelCapsule

Native CUDA layer normalization.

def NN.Backend.NativeCUDA.batchNorm : KernelCapsule

Native CUDA batch normalization.

def NN.Backend.NativeCUDA.batchNormChannelFirst : KernelCapsule

Native CUDA channel-first BatchNorm runtime op.

def NN.Backend.NativeCUDA.conv : KernelCapsule

Native CUDA generic channel-first convolution.

def NN.Backend.NativeCUDA.conv2d : KernelCapsule

Native CUDA 2D convolution.

def NN.Backend.NativeCUDA.convTranspose : KernelCapsule

Native CUDA generic channel-first transpose convolution.

def NN.Backend.NativeCUDA.convTranspose2d : KernelCapsule

Native CUDA 2D transpose convolution.

def NN.Backend.NativeCUDA.maxPool : KernelCapsule

Native CUDA max pooling.

def NN.Backend.NativeCUDA.maxPool2d : KernelCapsule

Native CUDA 2D max pooling.

def NN.Backend.NativeCUDA.maxPool2dPad : KernelCapsule

Native CUDA padded 2D max pooling.

def NN.Backend.NativeCUDA.smoothMaxPool : KernelCapsule

Native CUDA smooth max pooling.

def NN.Backend.NativeCUDA.smoothMaxPool2d : KernelCapsule

Native CUDA smooth 2D max pooling.

def NN.Backend.NativeCUDA.avgPool : KernelCapsule

Native CUDA average pooling.

def NN.Backend.NativeCUDA.avgPool2d : KernelCapsule

Native CUDA 2D average pooling.

def NN.Backend.NativeCUDA.avgPool2dPad : KernelCapsule

Native CUDA padded 2D average pooling.

def NN.Backend.NativeCUDA.linear : KernelCapsule

Native CUDA linear layer.

def NN.Backend.NativeCUDA.mseLoss : KernelCapsule

Native CUDA mean-squared-error loss.

def NN.Backend.NativeCUDA.fftFno : KernelCapsule

Native CUDA FFT/FNO kernels.

def NN.Backend.NativeCUDA.selectiveScan : KernelCapsule

Native CUDA selective scan kernels.

def NN.Backend.NativeCUDA.capsules : List KernelCapsule

Native CUDA capsules, excluding attention which has a dedicated semantic split.