make a few changes the import from transformers

modeling_gpt2.py

@@ -27,17 +27,17 @@ from torch import nn
 from torch.cuda.amp import autocast
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
 
-from ...activations import ACT2FN
-from ...modeling_outputs import (
+from transformers.activations import ACT2FN
+from transformers.modeling_outputs import (
     BaseModelOutputWithPastAndCrossAttentions,
     CausalLMOutputWithCrossAttentions,
     QuestionAnsweringModelOutput,
     SequenceClassifierOutputWithPast,
     TokenClassifierOutput,
 )
-from ...modeling_utils import PreTrainedModel, SequenceSummary
-from ...pytorch_utils import Conv1D, find_pruneable_heads_and_indices, prune_conv1d_layer
-from ...utils import (
+from transformers.modeling_utils import PreTrainedModel, SequenceSummary
+from transformers.pytorch_utils import Conv1D, find_pruneable_heads_and_indices, prune_conv1d_layer
+from transformers.utils import (
     ModelOutput,
     add_code_sample_docstrings,
     add_start_docstrings,
@@ -45,7 +45,7 @@ from ...utils import (
     logging,
     replace_return_docstrings,
 )
-from ...utils.model_parallel_utils import assert_device_map, get_device_map
+from transformers.utils.model_parallel_utils import assert_device_map, get_device_map
 from .configuration_gpt2 import GPT2Config