[tests] remove TF tests (uses of require_tf)

docs/source/de/testing.md

@@ -473,13 +473,6 @@ Hier ist zum Beispiel ein Test, der nur ausgeführt werden muss, wenn 2 oder meh
 def test_example_with_multi_gpu():
 ```
 
-Wenn ein Test `tensorflow` benötigt, verwenden Sie den Dekorator `require_tf`. Zum Beispiel:
-
-```python no-style
-@require_tf
-def test_tf_thing_with_tensorflow():
-```
-
 Diese Dekors können gestapelt werden. Wenn zum Beispiel ein Test langsam ist und mindestens eine GPU unter pytorch benötigt, können Sie
 wie Sie ihn einrichten können:
 
@@ -1204,9 +1197,6 @@ if torch.cuda.is_available():
 import numpy as np
 
 np.random.seed(seed)
-
-# tf RNG
-tf.random.set_seed(seed)
 ```
 
 ### Tests debuggen

docs/source/en/testing.md

@@ -474,13 +474,6 @@ For example, here is a test that must be run only when there are 2 or more GPUs
 def test_example_with_multi_gpu():
 ```
 
-If a test requires `tensorflow` use the `require_tf` decorator. For example:
-
-```python no-style
-@require_tf
-def test_tf_thing_with_tensorflow():
-```
-
 These decorators can be stacked. For example, if a test is slow and requires at least one GPU under pytorch, here is
 how to set it up:
 
@@ -1226,11 +1219,6 @@ if torch.cuda.is_available():
 import numpy as np
 
 np.random.seed(seed)
-
-# tf RNG
-import tensorflow as tf 
-
-tf.random.set_seed(seed)
 ```
 
 ### Debugging tests

docs/source/ja/testing.md

@@ -445,13 +445,6 @@ CUDA_VISIBLE_DEVICES="1" pytest tests/utils/test_logging.py
 def test_example_with_multi_gpu():
 ```
 
-テストに `tensorflow` が必要な場合は、`require_tf` デコレータを使用します。例えば：
-
-```python no-style
-@require_tf
-def test_tf_thing_with_tensorflow():
-```
-
 これらのデコレータは積み重ねることができます。たとえば、テストが遅く、pytorch で少なくとも 1 つの GPU が必要な場合は、次のようになります。
 設定方法:
 
@@ -1135,9 +1128,6 @@ if torch.cuda.is_available():
 import numpy as np
 
 np.random.seed(seed)
-
-# tf RNG
-tf.random.set_seed(seed)
 ```
 
 

docs/source/ko/testing.md

@@ -473,13 +473,6 @@ GPU 요구 사항을 표로 정리하면 아래와 같습니디ㅏ:
 def test_example_with_multi_gpu():
 ```
 
-`tensorflow`가 필요한 경우 `require_tf` 데코레이터를 사용합니다. 예를 들어 다음과 같습니다:
-
-```python no-style
-@require_tf
-def test_tf_thing_with_tensorflow():
-```
-
 이러한 데코레이터는 중첩될 수 있습니다.
 예를 들어, 느린 테스트로 진행되고 pytorch에서 적어도 하나의 GPU가 필요한 경우 다음과 같이 설정할 수 있습니다:
 

src/transformers/testing_utils.py

@@ -684,6 +684,9 @@ def require_tf(test_case):
     """
     Decorator marking a test that requires TensorFlow. These tests are skipped when TensorFlow isn't installed.
     """
+    logger.warning_once(
+        "TensorFlow test-related code, including `require_tf`, is deprecated and will be removed in Transformers v4.55"
+    )
     return unittest.skipUnless(is_tf_available(), "test requires TensorFlow")(test_case)
 
 

tests/models/layoutlmv3/test_tokenization_layoutlmv3.py

@@ -34,7 +34,6 @@
 from transformers.models.layoutlmv3.tokenization_layoutlmv3 import VOCAB_FILES_NAMES, LayoutLMv3Tokenizer
 from transformers.testing_utils import (
     require_pandas,
-    require_tf,
     require_tokenizers,
     require_torch,
     slow,
@@ -2306,42 +2305,6 @@ def test_layoutlmv3_integration_test(self):
     def test_np_encode_plus_sent_to_model(self):
         pass
 
-    @require_tf
-    @slow
-    def test_tf_encode_plus_sent_to_model(self):
-        from transformers import TF_MODEL_MAPPING, TOKENIZER_MAPPING
-
-        MODEL_TOKENIZER_MAPPING = merge_model_tokenizer_mappings(TF_MODEL_MAPPING, TOKENIZER_MAPPING)
-
-        tokenizers = self.get_tokenizers(do_lower_case=False)
-        for tokenizer in tokenizers:
-            with self.subTest(f"{tokenizer.__class__.__name__}"):
-                if tokenizer.__class__ not in MODEL_TOKENIZER_MAPPING:
-                    self.skipTest(f"{tokenizer.__class__} is not in the MODEL_TOKENIZER_MAPPING")
-
-                config_class, model_class = MODEL_TOKENIZER_MAPPING[tokenizer.__class__]
-                config = config_class()
-
-                if config.is_encoder_decoder or config.pad_token_id is None:
-                    self.skipTest(reason="Model is an encoder-decoder or has no pad token id set.")
-
-                model = model_class(config)
-
-                # Make sure the model contains at least the full vocabulary size in its embedding matrix
-                self.assertGreaterEqual(model.config.vocab_size, len(tokenizer))
-
-                # Build sequence
-                first_ten_tokens = list(tokenizer.get_vocab().keys())[:10]
-                boxes = [[1000, 1000, 1000, 1000] for _ in range(len(first_ten_tokens))]
-                encoded_sequence = tokenizer.encode_plus(first_ten_tokens, boxes=boxes, return_tensors="tf")
-                batch_encoded_sequence = tokenizer.batch_encode_plus(
-                    [first_ten_tokens, first_ten_tokens], boxes=[boxes, boxes], return_tensors="tf"
-                )
-
-                # This should not fail
-                model(encoded_sequence)
-                model(batch_encoded_sequence)
-
     @unittest.skip(reason="Chat is not supported")
     def test_chat_template(self):
         pass