src.tests.vl3dpp_dl_hierarchical_sg_pre_proc_test

Classes

VL3DPPDLHierarchicalSGPreProcTest()

class src.tests.vl3dpp_dl_hierarchical_sg_pre_proc_test.VL3DPPDLHierarchicalSGPreProcTest
Author:

Alberto M. Esmoris Pena

Deep learning pre-processor test that checks the C++ implementation of the hierarchical sparse grid (SG) pre-processors for deep learning models.

__init__()

Basic configuration for any VL3D test.

Parameters:

name (str) – Test name

run()

Run C++ deep learning hierarchical SG pre-processors test.

Returns:

True if the C++ pre-processors work as expected for the test cases, False otherwise.

Return type:

bool

validateHSGPreProc(hsgpre, hsgpre_ref, inputs)

Check that the hierarchical sparse grid pre-processor works as expected.

Parameters:

  • hsgpre – The object representing the Python-side pre-processor.

  • hsgpre_ref – The object representing the reference values.

  • inputs (dict) – The inputs for the pre-processor.

Returns:

True if the hierarchical sparse grid pre-processor worked as expected, False otherwise.

Return type:

bool

validateNxGuard(hsgpre)

Verify that the C++ pre-processor’s nx != 3 fast-fail (DLHierarchicalSGPreProcessor.tpp:89) fires when the coordinate matrix has a column count other than 3.

The guard rejects non-3D input before any of the nz = nt[2] / nynz = nt[1] * nt[2] 3D-specific stride reads can run on out-of-range memory. A regression in the guard would manifest at scale as silently miscomputed neighbor tables; this Python-side gate catches it directly.