styling

src/transformers/models/bark/modeling_bark.py

@@ -35,7 +35,6 @@
     add_start_docstrings_to_model_forward,
     is_accelerate_available,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
 )
 from ..auto import AutoModel

src/transformers/models/bart/modeling_bart.py

@@ -20,7 +20,6 @@
 from typing import List, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
@@ -48,7 +47,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/cohere/modeling_cohere.py

@@ -26,7 +26,6 @@
 from typing import List, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import CrossEntropyLoss
@@ -44,7 +43,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/data2vec/modeling_data2vec_audio.py

@@ -20,7 +20,6 @@
 
 import numpy as np
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import CrossEntropyLoss
@@ -41,7 +40,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     is_peft_available,
     logging,
 )

src/transformers/models/dbrx/modeling_dbrx.py

@@ -18,7 +18,6 @@
 from typing import Any, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 

src/transformers/models/distilbert/modeling_distilbert.py

@@ -23,7 +23,6 @@
 
 import numpy as np
 import torch
-import torch.nn.functional as F
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
 
@@ -45,7 +44,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/falcon/modeling_falcon.py

@@ -43,7 +43,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
 )
 from .configuration_falcon import FalconConfig

src/transformers/models/gpt2/modeling_gpt2.py

@@ -22,7 +22,6 @@
 from typing import Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
@@ -52,8 +51,7 @@
 
 
 if is_flash_attn_2_available():
-    from flash_attn import flash_attn_func, flash_attn_varlen_func
-    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input
+    from ...flash_attention_utils import _flash_attention_forward
 
 
 logger = logging.get_logger(__name__)
@@ -62,9 +60,6 @@
 _CONFIG_FOR_DOC = "GPT2Config"
 
 
-
-
-
 def load_tf_weights_in_gpt2(model, config, gpt2_checkpoint_path):
     """Load tf checkpoints in a pytorch model"""
     try:

src/transformers/models/gpt_bigcode/modeling_gpt_bigcode.py

@@ -17,7 +17,6 @@
 from typing import List, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
@@ -37,7 +36,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
 )
 from .configuration_gpt_bigcode import GPTBigCodeConfig

src/transformers/models/gpt_neo/modeling_gpt_neo.py

@@ -18,7 +18,6 @@
 from typing import Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
@@ -41,7 +40,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     is_torch_fx_available,
     logging,
 )

src/transformers/models/gptj/modeling_gptj.py

@@ -19,7 +19,6 @@
 
 import torch
 import torch.fx
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss

src/transformers/models/hubert/modeling_hubert.py

@@ -19,7 +19,6 @@
 
 import numpy as np
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import CrossEntropyLoss
@@ -33,7 +32,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/idefics2/modeling_idefics2.py

@@ -14,13 +14,11 @@
 # limitations under the License.
 """PyTorch Idefics2 model."""
 
-import inspect
 import math
 from dataclasses import dataclass
 from typing import Dict, List, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import CrossEntropyLoss
@@ -34,7 +32,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/jamba/modeling_jamba.py

@@ -19,7 +19,6 @@
 # limitations under the License.
 """PyTorch Jamba model."""
 
-import inspect
 import math
 from typing import Any, Dict, List, Optional, Tuple, Union
 
@@ -43,7 +42,6 @@
 from ...utils import (
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )
@@ -58,8 +56,6 @@
 if is_flash_attn_2_available():
     from ...flash_attention_utils import _flash_attention_forward
 
-    _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)
-
 
 if is_mamba_ssm_available():
     from mamba_ssm.ops.selective_scan_interface import mamba_inner_fn, selective_scan_fn
@@ -478,7 +474,7 @@ def forward(
             q_len,
             dropout=dropout_rate,
             sliding_window=getattr(self.config, "sliding_window", None) ,
-            cache_position=cache_position 
+            cache_position=cache_position
         )
 
         attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()

src/transformers/models/m2m_100/modeling_m2m_100.py

@@ -18,7 +18,6 @@
 from typing import List, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 from torch import nn
 from torch.nn import CrossEntropyLoss
 
@@ -46,8 +45,7 @@
 
 
 if is_flash_attn_2_available():
-    from flash_attn import flash_attn_func, flash_attn_varlen_func
-    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input
+    from ...flash_attention_utils import _flash_attention_forward
 
 
 logger = logging.get_logger(__name__)

src/transformers/models/mbart/modeling_mbart.py

@@ -19,7 +19,6 @@
 from typing import List, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
@@ -42,7 +41,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/mistral/modeling_mistral.py

@@ -41,7 +41,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/mixtral/modeling_mixtral.py

@@ -19,7 +19,6 @@
 # limitations under the License.
 """PyTorch Mixtral model."""
 
-import inspect
 import math
 from typing import List, Optional, Tuple, Union
 
@@ -47,7 +46,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )
@@ -58,8 +56,6 @@
 if is_flash_attn_2_available():
     from ...flash_attention_utils import _flash_attention_forward
 
-    _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)
-
 # This makes `_prepare_4d_causal_attention_mask` a leaf function in the FX graph.
 # It means that the function will not be traced through and simply appear as a node in the graph.
 if is_torch_fx_available():

src/transformers/models/musicgen/modeling_musicgen.py

@@ -23,7 +23,6 @@
 
 import torch
 import torch.nn as nn
-import torch.nn.functional as F
 from torch.nn import CrossEntropyLoss
 
 from ...activations import ACT2FN
@@ -48,7 +47,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/musicgen_melody/modeling_musicgen_melody.py

@@ -23,7 +23,6 @@
 
 import torch
 import torch.nn as nn
-import torch.nn.functional as F
 from torch.nn import CrossEntropyLoss
 
 from ...activations import ACT2FN
@@ -40,7 +39,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/olmo/modeling_olmo.py

@@ -41,7 +41,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/opt/modeling_opt.py

@@ -17,7 +17,6 @@
 from typing import List, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
@@ -36,7 +35,6 @@
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )

src/transformers/models/phi/modeling_phi.py

@@ -19,7 +19,6 @@
 from typing import List, Optional, Tuple, Union
 
 import torch
-import torch.nn.functional as F
 import torch.utils.checkpoint
 from packaging import version
 from torch import nn
@@ -44,7 +43,6 @@
     add_start_docstrings_to_model_forward,
     get_torch_version,
     is_flash_attn_2_available,
-    is_flash_attn_greater_or_equal_2_10,
     logging,
     replace_return_docstrings,
 )