dev(narugo): test transformers clip with real cases

from unittest import skipUnless

import numpy as np

import pytest

from hbutils.testing import tmatrix

from imgutils.data import load_image

from imgutils.preprocess.transformers import create_transforms_from_transformers

from test.testings import get_testfile

try:

    import transformers

except (ImportError, ModuleNotFoundError):

    _HAS_TRANSFORMERS = False

else:

    _HAS_TRANSFORMERS = True

@pytest.mark.unittest

class TestPreprocessTransformersAlign:

    @skipUnless(_HAS_TRANSFORMERS, 'Transformers required.')

    @pytest.mark.parametrize(*tmatrix({

        'repo_id': [

            "openai/clip-vit-base-patch32",

            "openai/clip-vit-large-patch14",

            "openai/clip-vit-large-patch14-336",

            "microsoft/Florence-2-large",

            "llava-hf/llava-1.5-7b-hf",

            'facebook/metaclip-b32-400m',

            'apple/aimv2-large-patch14-224',

            'nvidia/RADIO',

        ],

        'src_image': [

            'png_640.png',

            'png_640_m90.png',

            'nude_girl.png',

            'dori_640.png',

            'nian_640.png',

        ]

    }))

    def test_image_preprocess_align(self, src_image, repo_id):

        from transformers import AutoImageProcessor

        image = load_image(get_testfile(src_image), mode='RGB', force_background='white')

        processor = AutoImageProcessor.from_pretrained(repo_id)

        trans = create_transforms_from_transformers(processor)

        expected_output = processor.preprocess(image)['pixel_values'][0]

        output = trans(image)

        np.testing.assert_array_almost_equal(

            output,

            expected_output,

        )

    @skipUnless(_HAS_TRANSFORMERS, 'Transformers required.')

    @pytest.mark.parametrize(*tmatrix({

        'repo_id': [

            "openai/clip-vit-base-patch32",

            "openai/clip-vit-large-patch14",

            "openai/clip-vit-large-patch14-336",

        ],

        'src_image': [

            'png_640.png',

            'png_640_m90.png',

            'nude_girl.png',

            'dori_640.png',

            'nian_640.png',

        ]

    }))

    def test_auto_preprocess_align(self, src_image, repo_id):

        from transformers import AutoProcessor

        image = load_image(get_testfile(src_image), mode='RGB', force_background='white')

        processor = AutoProcessor.from_pretrained(repo_id)

        trans = create_transforms_from_transformers(processor)

        expected_output = processor.image_processor.preprocess(image)['pixel_values'][0]

        output = trans(image)

        np.testing.assert_array_almost_equal(

            output,

            expected_output,

        )