Fix typos (#37978)

fix typos

tests/models/colpali/test_modeling_colpali.py

@@ -275,7 +275,7 @@ class ColPaliForRetrievalModelTest(ModelTesterMixin, unittest.TestCase):
         pass
 
     @unittest.skip(
-        reason="PaliGemmma's SigLip encoder uses the same initialization scheme as the Flax original implementation"
+        reason="PaliGemma's SigLip encoder uses the same initialization scheme as the Flax original implementation"
     )
     def test_initialization(self):
         pass

tests/models/deepseek_v3/test_modeling_deepseek_v3.py

@@ -431,7 +431,7 @@ class DeepseekV3ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTeste
 
     def test_past_key_values_format(self):
         """
-        Overwritting to pass the expected cache shapes (Deepseek-V3 uses MLA so the cache shapes are non-standard)
+        Overwriting to pass the expected cache shapes (Deepseek-V3 uses MLA so the cache shapes are non-standard)
         """
         config, inputs = self.model_tester.prepare_config_and_inputs_for_common()
         batch_size, seq_length = inputs["input_ids"].shape
@@ -451,7 +451,7 @@ class DeepseekV3ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTeste
     @slow
     def test_eager_matches_sdpa_generate(self):
         """
-        Overwritting the common test as the test is flaky on tiny models
+        Overwriting the common test as the test is flaky on tiny models
         """
         max_new_tokens = 30
 

tests/models/marian/test_tokenization_marian.py

@@ -136,7 +136,7 @@ class MarianTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
             decode_kwargs={"use_source_tokenizer": True},
         )
 
-    def test_tokenizer_integration_seperate_vocabs(self):
+    def test_tokenizer_integration_separate_vocabs(self):
         tokenizer = MarianTokenizer.from_pretrained("hf-internal-testing/test-marian-two-vocabs")
 
         source_text = "Tämä on testi"

tests/models/opt/test_modeling_flax_opt.py

@@ -69,7 +69,7 @@ class FlaxOPTModelTester:
         embed_dim=16,
         word_embed_proj_dim=16,
         initializer_range=0.02,
-        attn_implemetation="eager",
+        attn_implementation="eager",
     ):
         self.parent = parent
         self.batch_size = batch_size
@@ -92,7 +92,7 @@ class FlaxOPTModelTester:
         self.word_embed_proj_dim = word_embed_proj_dim
         self.initializer_range = initializer_range
         self.is_encoder_decoder = False
-        self.attn_implementation = attn_implemetation
+        self.attn_implementation = attn_implementation
 
     def prepare_config_and_inputs(self):
         input_ids = np.clip(ids_tensor([self.batch_size, self.seq_length - 1], self.vocab_size), 3, self.vocab_size)

tests/models/paligemma/test_modeling_paligemma.py

@@ -297,7 +297,7 @@ class PaliGemmaForConditionalGenerationModelTest(ModelTesterMixin, GenerationTes
         pass
 
     @unittest.skip(
-        reason="PaliGemmma's SigLip encoder uses the same initialization scheme as the Flax original implementation"
+        reason="PaliGemma's SigLip encoder uses the same initialization scheme as the Flax original implementation"
     )
     def test_initialization(self):
         pass

tests/models/paligemma2/test_modeling_paligemma2.py

@@ -294,7 +294,7 @@ class PaliGemma2ForConditionalGenerationModelTest(ModelTesterMixin, GenerationTe
         pass
 
     @unittest.skip(
-        reason="PaliGemmma's SigLip encoder uses the same initialization scheme as the Flax original implementation"
+        reason="PaliGemma's SigLip encoder uses the same initialization scheme as the Flax original implementation"
     )
     def test_initialization(self):
         pass