[Good first issue] MPNet TensorFlow Integration tests

[LysandreJik]

The TensorFlow implementation of the MPNet model currently has no integration tests. This is problematic as the behavior can diverge without being noticed.

The test_modeling_tf_mpnet.py file should be updated to include integration testing.

An example of a good modeling integration test is visible in the test_modeling_tf_bert.py#L365-L387 file:

transformers/tests/test_modeling_tf_bert.py

Lines 365 to 387 in 1809de5

	@require_tf
	class TFBertModelIntegrationTest(unittest.TestCase):
	@slow
	def test_inference_masked_lm(self):
	model = TFBertForPreTraining.from_pretrained("lysandre/tiny-bert-random")
	input_ids = tf.constant([[0, 1, 2, 3, 4, 5]])
	output = model(input_ids)[0]
	expected_shape = [1, 6, 32000]
	self.assertEqual(output.shape, expected_shape)
	print(output[:, :3, :3])
	expected_slice = tf.constant(
	[
	[
	[-0.05243197, -0.04498899, 0.05512108],
	[-0.07444685, -0.01064632, 0.04352357],
	[-0.05020351, 0.05530146, 0.00700043],
	]
	]
	)
	tf.debugging.assert_near(output[:, :3, :3], expected_slice, atol=1e-4)

Some additional tips:

• The test must be marked as slow using the @slow decorator, so as to be run daily, and not on every commit of every branch/pull request of this repository.
• The test must be decorated with the @require_tf decorator so as to only be run in environments using PyTorch.
• A single test is necessary. If you feel like implementing multiple of these, then sharing the same checkpoint would be ideal so as to reduce download time.