fix: Removed duplicate field definitions in some classes

examples/flax/language-modeling/run_clm_flax.py

@@ -225,9 +225,6 @@ class DataTrainingArguments:
             )
         },
     )
-    overwrite_cache: bool = field(
-        default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
-    )
     validation_split_percentage: Optional[int] = field(
         default=5,
         metadata={

examples/research_projects/jax-projects/hybrid_clip/run_hybrid_clip.py

@@ -163,9 +163,6 @@ class DataTrainingArguments:
     overwrite_cache: bool = field(
         default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
     )
-    overwrite_cache: bool = field(
-        default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
-    )
     preprocessing_num_workers: Optional[int] = field(
         default=None,
         metadata={"help": "The number of processes to use for the preprocessing."},

examples/research_projects/jax-projects/model_parallel/run_clm_mp.py

@@ -156,9 +156,6 @@ class DataTrainingArguments:
             )
         },
     )
-    overwrite_cache: bool = field(
-        default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
-    )
     preprocessing_num_workers: Optional[int] = field(
         default=None,
         metadata={"help": "The number of processes to use for the preprocessing."},

src/transformers/models/deformable_detr/modeling_deformable_detr.py

@@ -1080,7 +1080,6 @@ class DeformableDetrPreTrainedModel(PreTrainedModel):
     main_input_name = "pixel_values"
     supports_gradient_checkpointing = True
     _no_split_modules = [r"DeformableDetrConvEncoder", r"DeformableDetrEncoderLayer", r"DeformableDetrDecoderLayer"]
-    supports_gradient_checkpointing = True
 
     def _init_weights(self, module):
         std = self.config.init_std

src/transformers/models/video_llava/modeling_video_llava.py

@@ -126,7 +126,6 @@ class VideoLlavaPreTrainedModel(PreTrainedModel):
     _no_split_modules = ["VideoLlavaVisionAttention"]
     _skip_keys_device_placement = "past_key_values"
     _supports_flash_attn_2 = True
-    _no_split_modules = ["VideoLlavaVisionAttention"]
 
     def _init_weights(self, module):
         std = (

tests/models/fnet/test_modeling_fnet.py

@@ -295,7 +295,6 @@ class FNetModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
     # Skip Tests
     test_pruning = False
     test_head_masking = False
-    test_pruning = False
 
     # TODO: Fix the failed tests
     def is_pipeline_test_to_skip(

tests/models/mamba/test_modeling_mamba.py

@@ -258,7 +258,6 @@ class MambaModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixi
     test_model_parallel = False
     test_pruning = False
     test_head_masking = False  # Mamba does not have attention heads
-    test_model_parallel = False
     pipeline_model_mapping = (
         {"feature-extraction": MambaModel, "text-generation": MambaForCausalLM} if is_torch_available() else {}
     )

tests/models/recurrent_gemma/test_modeling_recurrent_gemma.py

@@ -298,7 +298,6 @@ class RecurrentGemmaModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineT
     test_model_parallel = False
     test_pruning = False
     test_head_masking = False  # RecurrentGemma does not have attention heads
-    test_model_parallel = False
 
     # Need to remove 0.9 in `test_cpu_offload`
     # This is because we are hitting edge cases with the causal_mask buffer