zhihan1996
commited on
Commit
•
1ce72a5
1
Parent(s):
1c51ecc
Update bert_layers.py
Browse files- bert_layers.py +4 -3
bert_layers.py
CHANGED
@@ -18,6 +18,7 @@ from torch.nn.modules.utils import consume_prefix_in_state_dict_if_present
|
|
18 |
from transformers.activations import ACT2FN
|
19 |
from transformers.modeling_outputs import (MaskedLMOutput,
|
20 |
SequenceClassifierOutput)
|
|
|
21 |
from transformers.modeling_utils import PreTrainedModel
|
22 |
|
23 |
from .bert_padding import (index_first_axis,
|
@@ -521,7 +522,7 @@ class BertPredictionHeadTransform(nn.Module):
|
|
521 |
return hidden_states
|
522 |
|
523 |
|
524 |
-
class BertModel(
|
525 |
"""Overall BERT model.
|
526 |
|
527 |
Args:
|
@@ -681,7 +682,7 @@ class BertOnlyNSPHead(nn.Module):
|
|
681 |
|
682 |
|
683 |
|
684 |
-
class BertForMaskedLM(
|
685 |
|
686 |
def __init__(self, config):
|
687 |
super().__init__(config)
|
@@ -810,7 +811,7 @@ class BertForMaskedLM(PreTrainedModel):
|
|
810 |
|
811 |
|
812 |
|
813 |
-
class BertForSequenceClassification(
|
814 |
"""Bert Model transformer with a sequence classification/regression head.
|
815 |
|
816 |
This head is just a linear layer on top of the pooled output. Used for,
|
|
|
18 |
from transformers.activations import ACT2FN
|
19 |
from transformers.modeling_outputs import (MaskedLMOutput,
|
20 |
SequenceClassifierOutput)
|
21 |
+
from transformers.modeling_bert import BertPreTrainedModel
|
22 |
from transformers.modeling_utils import PreTrainedModel
|
23 |
|
24 |
from .bert_padding import (index_first_axis,
|
|
|
522 |
return hidden_states
|
523 |
|
524 |
|
525 |
+
class BertModel(BertPreTrainedModel):
|
526 |
"""Overall BERT model.
|
527 |
|
528 |
Args:
|
|
|
682 |
|
683 |
|
684 |
|
685 |
+
class BertForMaskedLM(BertPreTrainedModel):
|
686 |
|
687 |
def __init__(self, config):
|
688 |
super().__init__(config)
|
|
|
811 |
|
812 |
|
813 |
|
814 |
+
class BertForSequenceClassification(BertPreTrainedModel):
|
815 |
"""Bert Model transformer with a sequence classification/regression head.
|
816 |
|
817 |
This head is just a linear layer on top of the pooled output. Used for,
|