skainet-compile-hlo/sk.ainet.compile.hlo.examples/ShapeOperationsExample

ShapeOperationsExample

object ShapeOperationsExample(source)

Example demonstrating shape manipulation operations in StableHLO conversion.

This example shows how to:

Create a graph with various shape operations
Convert reshape, flatten, squeeze, and unsqueeze operations to StableHLO
Handle dynamic shape transformations

Functions

convertToStableHlo

fun convertToStableHlo(graph: DefaultComputeGraph, functionName: String = "shape_ops"): String

Convert a graph to StableHLO MLIR using the extended converter

createNeuralNetworkFlattenGraph

fun createNeuralNetworkFlattenGraph(): DefaultComputeGraph

Create a graph demonstrating flatten operation for neural networks: input (1,32,28,28) -> flatten (1,25088) typical CNN to FC transition

createShapeTransformationGraph

fun createShapeTransformationGraph(): DefaultComputeGraph

Create a graph demonstrating common shape transformations: input (2,3,4,5) -> reshape (2,3,20) -> flatten (2,60) -> unsqueeze (2,1,60) -> squeeze (2,60)

createSimpleReshapeGraph

fun createSimpleReshapeGraph(): DefaultComputeGraph

Create a simpler graph showing just reshape operation: input (4,6) -> reshape (2,12)

runComplexExample

fun runComplexExample(): String

Run the complete shape transformation example

runNeuralNetworkExample

fun runNeuralNetworkExample(): String

Run the neural network flatten example

runSimpleExample

fun runSimpleExample(): String

Run the simple reshape example