Fix TF nightly tests (#20507)

* Fixed test_saved_model_extended

* Fix TFGPT2 tests

* make fixup

* Make sure keras-nlp utils are available for type hinting too

* Update src/transformers/testing_utils.py
Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

* make fixup
Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

src/transformers/__init__.py

@@ -525,6 +525,7 @@ _import_structure = {
         "is_datasets_available",
         "is_faiss_available",
         "is_flax_available",
+        "is_keras_nlp_available",
         "is_phonemizer_available",
         "is_psutil_available",
         "is_py3nvml_available",
@@ -3706,6 +3707,7 @@ if TYPE_CHECKING:
         is_datasets_available,
         is_faiss_available,
         is_flax_available,
+        is_keras_nlp_available,
         is_phonemizer_available,
         is_psutil_available,
         is_py3nvml_available,

src/transformers/testing_utils.py

@@ -57,6 +57,7 @@ from .utils import (
     is_ftfy_available,
     is_ipex_available,
     is_jumanpp_available,
+    is_keras_nlp_available,
     is_librosa_available,
     is_natten_available,
     is_onnx_available,
@@ -392,6 +393,13 @@ def require_tensorflow_text(test_case):
     return unittest.skipUnless(is_tensorflow_text_available(), "test requires tensorflow_text")(test_case)
 
 
+def require_keras_nlp(test_case):
+    """
+    Decorator marking a test that requires keras_nlp. These tests are skipped when keras_nlp isn't installed.
+    """
+    return unittest.skipUnless(is_keras_nlp_available(), "test requires keras_nlp")(test_case)
+
+
 def require_pandas(test_case):
     """
     Decorator marking a test that requires pandas. These tests are skipped when pandas isn't installed.

tests/models/gpt2/test_tokenization_gpt2_tf.py

@@ -2,12 +2,12 @@ import unittest
 from pathlib import Path
 from tempfile import TemporaryDirectory
 
-from transformers import AutoConfig, TFGPT2LMHeadModel, is_tensorflow_text_available, is_tf_available
+from transformers import AutoConfig, TFGPT2LMHeadModel, is_keras_nlp_available, is_tf_available
 from transformers.models.gpt2.tokenization_gpt2 import GPT2Tokenizer
-from transformers.testing_utils import require_tensorflow_text, slow
+from transformers.testing_utils import require_keras_nlp, slow
 
 
-if is_tensorflow_text_available():
+if is_keras_nlp_available():
     from transformers.models.gpt2 import TFGPT2Tokenizer
 
 if is_tf_available():
@@ -40,7 +40,7 @@ if is_tf_available():
             return outputs
 
 
-@require_tensorflow_text
+@require_keras_nlp
 class GPTTokenizationTest(unittest.TestCase):
     # The TF tokenizers are usually going to be used as pretrained tokenizers from existing model checkpoints,
     # so that's what we focus on here.

tests/utils/test_modeling_tf_core.py

@@ -218,11 +218,17 @@ class TFCoreModelTesterMixin:
             model = model_class(config)
             num_out = len(model(class_inputs_dict))
 
-            for key in class_inputs_dict.keys():
+            for key in list(class_inputs_dict.keys()):
+                # Remove keys not in the serving signature, as the SavedModel will not be compiled to deal with them
+                if key not in model.serving.input_signature[0]:
+                    del class_inputs_dict[key]
                 # Check it's a tensor, in case the inputs dict has some bools in it too
-                if isinstance(class_inputs_dict[key], tf.Tensor) and class_inputs_dict[key].dtype.is_integer:
+                elif isinstance(class_inputs_dict[key], tf.Tensor) and class_inputs_dict[key].dtype.is_integer:
                     class_inputs_dict[key] = tf.cast(class_inputs_dict[key], tf.int32)
 
+            if set(class_inputs_dict.keys()) != set(model.serving.input_signature[0].keys()):
+                continue  # Some models have inputs that the preparation functions don't create, we skip those
+
             with tempfile.TemporaryDirectory() as tmpdirname:
                 model.save_pretrained(tmpdirname, saved_model=True)
                 saved_model_dir = os.path.join(tmpdirname, "saved_model", "1")