test/utils/test_memory.py

import torch
from torch.autograd import gradcheck

import kornia
import kornia.testing as utils


class TestBatchedForward:
    def test_runbatch(self, device):
        patches = torch.rand(34, 1, 32, 32)
        sift = kornia.feature.SIFTDescriptor(32)
        desc_batched = kornia.utils.memory.batched_forward(sift, patches, device, 32)
        desc = sift(patches)
        assert torch.allclose(desc, desc_batched)

    def test_runone(self, device):
        patches = torch.rand(16, 1, 32, 32)
        sift = kornia.feature.SIFTDescriptor(32)
        desc_batched = kornia.utils.memory.batched_forward(sift, patches, device, 32)
        desc = sift(patches)
        assert torch.allclose(desc, desc_batched)

    def test_gradcheck(self, device):
        batch_size, channels, height, width = 3, 2, 5, 4
        img = torch.rand(batch_size, channels, height, width, device=device)
        img = utils.tensor_to_gradcheck_var(img)  # to var
        assert gradcheck(
            kornia.utils.memory.batched_forward,
            (kornia.feature.BlobHessian(), img, device, 2),
            raise_exception=True,
            nondet_tol=1e-4,
        )