def test_inference_with_head(self):
        model = TFLEDForConditionalGeneration.from_pretrained("allenai/led-base-16384")

        # change to intended input here
        input_ids = _long_tensor([512 * [0, 31414, 232, 328, 740, 1140, 12695, 69]])
        decoder_input_ids = _long_tensor([128 * [0, 31414, 232, 328, 740, 1140, 12695, 69]])
        inputs_dict = prepare_led_inputs_dict(model.config, input_ids, decoder_input_ids)
        output = model(**inputs_dict)[0]
        expected_shape = (1, 1024, model.config.vocab_size)
        self.assertEqual(output.shape, expected_shape)
        # change to expected output here
        expected_slice = tf.convert_to_tensor(
            [[33.6507, 6.4572, 16.8089], [5.8739, -2.4238, 11.2902], [-3.2139, -4.3149, 4.2783]],
        )
        tf.debugging.assert_near(output[:, :3, :3], expected_slice, atol=TOLERANCE)
    def test_inference_no_head(self):
        model = TFLEDForConditionalGeneration.from_pretrained("allenai/led-base-16384").led

        # change to intended input here
        input_ids = _long_tensor([512 * [0, 31414, 232, 328, 740, 1140, 12695, 69]])
        decoder_input_ids = _long_tensor([128 * [0, 31414, 232, 328, 740, 1140, 12695, 69]])
        inputs_dict = prepare_led_inputs_dict(model.config, input_ids, decoder_input_ids)
        output = model(**inputs_dict)[0]
        expected_shape = (1, 1024, 768)
        self.assertEqual(output.shape, expected_shape)
        # change to expected output here
        expected_slice = tf.convert_to_tensor(
            [[2.3050, 2.8279, 0.6531], [-1.8457, -0.1455, -3.5661], [-1.0186, 0.4586, -2.2043]],
        )
        tf.debugging.assert_near(output[:, :3, :3], expected_slice, atol=TOLERANCE)