Method setUp

tests/input/test_ernie_vl_processor.py:902–940

Set up test environment

(self)

class TestDataProcessor(unittest.TestCase):
    def setUp(self):
        """Set up test environment"""
        self.mock_tokenizer = MagicMock()

        def mock_convert_tokens_to_ids(x):
            if isinstance(x, list):
                return [hash(str(token)) % 10000 for token in x]
            return hash(str(x)) % 10000

        self.mock_tokenizer.convert_tokens_to_ids = MagicMock(side_effect=mock_convert_tokens_to_ids)
        self.mock_tokenizer.encode = MagicMock(return_value={"input_ids": [1, 2, 3]})
        self.mock_tokenizer.decode = MagicMock(return_value="decoded_text")
        self.mock_tokenizer.tokenize = MagicMock(return_value=["token1", "token2"])
        self.mock_tokenizer.ignored_index = -100
        self.mock_tokenizer.chat_template = MagicMock()
        self.mock_tokenizer.apply_chat_template = MagicMock(return_value="formatted_prompt")

        self.mock_image_preprocessor = MagicMock()
        self.mock_image_preprocessor.get_smarted_resize = MagicMock(return_value=((224, 224), (16, 16)))
        self.mock_image_preprocessor.preprocess = MagicMock(
            return_value={
                "pixel_values": np.random.rand(256, 3 * 14 * 14).astype(np.float32),
                "image_grid_thw": np.array([[1, 16, 16]]),
            }
        )
        self.mock_image_preprocessor.from_pretrained = MagicMock(return_value=self.mock_image_preprocessor)

        with patch(
            "fastdeploy.input.ernie4_5_vl_processor.process.AdaptiveImageProcessor",
            self.mock_image_preprocessor,
        ):
            with patch("fastdeploy.input.ernie4_5_vl_processor.process.Ernie4_5Tokenizer") as mock_tokenizer_class:
                mock_tokenizer_class.from_pretrained = MagicMock(return_value=self.mock_tokenizer)
                mock_tokenizer_class.resource_files_names = {"vocab_file": "tokenizer.model"}
                with patch("os.path.exists", return_value=True):
                    self.processor = DataProcessor(
                        tokenizer_name="test_model",
                        image_preprocessor_name="test_model",
                    )

    def _create_outputs(self):
        """Helper to create outputs dict"""