Add an annotation to tests that need kaggle auth (#1470)

We can skip these by default, for users who have not yet set them up.
We will need to set them up for CI, see
keras-team/keras-hub#1459

keras_nlp/conftest.py

@@ -12,6 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import os
+
 import pytest
 import tensorflow as tf
 
@@ -83,6 +85,10 @@ def pytest_configure(config):
  "markers",
  "keras_3_only: mark test as a keras 3 only test",
  )
+ config.addinivalue_line(
+ "markers",
+ "kaggle_key_required: mark test needing a kaggle key",
+ )
 
 
 def pytest_collection_modifyitems(config, items):
@@ -107,6 +113,16 @@ def pytest_collection_modifyitems(config, items):
  not backend_config.keras_3(),
  reason="tests only run on with multi-backend keras",
  )
+ found_kaggle_key = all(
+ [
+ os.environ.get("KAGGLE_USERNAME", None),
+ os.environ.get("KAGGLE_KEY", None),
+ ]
+ )
+ kaggle_key_required = pytest.mark.skipif(
+ not found_kaggle_key,
+ reason="tests only run with a kaggle api key",
+ )
  for item in items:
  if "large" in item.keywords:
  item.add_marker(skip_large)
@@ -116,6 +132,8 @@ def pytest_collection_modifyitems(config, items):
  item.add_marker(tf_only)
  if "keras_3_only" in item.keywords:
  item.add_marker(keras_3_only)
+ if "kaggle_key_required" in item.keywords:
+ item.add_marker(kaggle_key_required)
 
 
 # Disable traceback filtering for quicker debugging of tests failures.

keras_nlp/models/gemma/gemma_backbone_test.py

@@ -53,6 +53,7 @@ def test_saved_model(self):
  input_data=self.input_data,
  )
 
+ @pytest.mark.kaggle_key_required
  @pytest.mark.large
  def test_smallest_preset(self):
  self.run_preset_test(
@@ -69,6 +70,7 @@ def test_smallest_preset(self):
  ),
  )
 
+ @pytest.mark.kaggle_key_required
  @pytest.mark.extra_large
  def test_all_presets(self):
  for preset in GemmaBackbone.presets:

keras_nlp/models/gemma/gemma_causal_lm_preprocessor_test.py

@@ -82,6 +82,7 @@ def test_generate_postprocess(self):
  x = preprocessor.generate_postprocess(input_data)
  self.assertAllEqual(x, "the quick brown fox")
 
+ @pytest.mark.kaggle_key_required
  @pytest.mark.extra_large
  def test_all_presets(self):
  for preset in GemmaCausalLMPreprocessor.presets:

keras_nlp/models/gemma/gemma_causal_lm_test.py

@@ -142,6 +142,7 @@ def test_generate_compilation(self):
  causal_lm.compile(sampler="greedy")
  self.assertIsNone(causal_lm.generate_function)
 
+ @pytest.mark.kaggle_key_required
  @pytest.mark.large
  def test_saved_model(self):
  self.run_model_saving_test(
@@ -150,6 +151,7 @@ def test_saved_model(self):
  input_data=self.input_data,
  )
 
+ @pytest.mark.kaggle_key_required
  @pytest.mark.extra_large
  def test_all_presets(self):
  for preset in GemmaCausalLM.presets:

keras_nlp/models/gemma/gemma_preprocessor_test.py

@@ -64,6 +64,7 @@ def test_sequence_length_override(self):
  x = preprocessor(input_data, sequence_length=4)
  self.assertAllEqual(x["token_ids"], [1, 4, 9, 2])
 
+ @pytest.mark.kaggle_key_required
  @pytest.mark.extra_large
  def test_all_presets(self):
  for preset in GemmaPreprocessor.presets:

keras_nlp/models/gemma/gemma_tokenizer_test.py

@@ -48,6 +48,7 @@ def test_errors_missing_special_tokens(self):
  )
  )
 
+ @pytest.mark.kaggle_key_required
  @pytest.mark.large
  def test_smallest_preset(self):
  self.run_preset_test(
@@ -57,6 +58,7 @@ def test_smallest_preset(self):
  expected_output=[[651, 4320, 8426, 25341, 235265]],
  )
 
+ @pytest.mark.kaggle_key_required
  @pytest.mark.extra_large
  def test_all_presets(self):
  for preset in GemmaTokenizer.presets: