kendallpark
commited on
Commit
•
f1f6ae0
1
Parent(s):
b59c09e
Update modeling_cxrbert.py
Browse files- modeling_cxrbert.py +45 -27
modeling_cxrbert.py
CHANGED
@@ -3,6 +3,7 @@
|
|
3 |
# Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
|
4 |
# ------------------------------------------------------------------------------------------
|
5 |
|
|
|
6 |
from typing import Any, Optional, Tuple, Union
|
7 |
|
8 |
import torch
|
@@ -16,20 +17,24 @@ from .configuration_cxrbert import CXRBertConfig
|
|
16 |
|
17 |
BERTTupleOutput = Tuple[T, T, T, T, T]
|
18 |
|
|
|
|
|
19 |
class CXRBertOutput(ModelOutput):
|
20 |
last_hidden_state: torch.FloatTensor
|
21 |
-
logits: torch.FloatTensor
|
22 |
cls_projected_embedding: Optional[torch.FloatTensor] = None
|
23 |
hidden_states: Optional[Tuple[torch.FloatTensor]] = None
|
24 |
attentions: Optional[Tuple[torch.FloatTensor]] = None
|
25 |
|
26 |
|
27 |
class BertProjectionHead(nn.Module):
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
|
|
|
|
33 |
def __init__(self, config: CXRBertConfig) -> None:
|
34 |
super().__init__()
|
35 |
self.dense_to_hidden = nn.Linear(config.hidden_size, config.projection_size)
|
@@ -50,13 +55,13 @@ class CXRBertModel(BertForMaskedLM):
|
|
50 |
"""
|
51 |
Implements the CXR-BERT model outlined in the manuscript:
|
52 |
Boecking et al. "Making the Most of Text Semantics to Improve Biomedical Vision-Language Processing", 2022
|
53 |
-
https://
|
54 |
|
55 |
-
Extends the HuggingFace BertForMaskedLM model by adding a separate projection head. The projection "[CLS]" token is
|
56 |
-
the latent vectors of image and text modalities.
|
57 |
"""
|
58 |
|
59 |
-
config_class = CXRBertConfig
|
60 |
|
61 |
def __init__(self, config: CXRBertConfig):
|
62 |
super().__init__(config)
|
@@ -78,21 +83,24 @@ class CXRBertModel(BertForMaskedLM):
|
|
78 |
return_dict: Optional[bool] = None,
|
79 |
**kwargs: Any
|
80 |
) -> Union[BERTTupleOutput, CXRBertOutput]:
|
81 |
-
|
82 |
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
83 |
|
84 |
-
bert_for_masked_lm_output = super().forward(
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
|
|
|
|
93 |
|
94 |
last_hidden_state = bert_for_masked_lm_output.hidden_states[-1]
|
95 |
-
cls_projected_embedding =
|
|
|
|
|
96 |
|
97 |
if return_dict:
|
98 |
return CXRBertOutput(
|
@@ -108,21 +116,31 @@ class CXRBertModel(BertForMaskedLM):
|
|
108 |
bert_for_masked_lm_output.logits,
|
109 |
cls_projected_embedding,
|
110 |
bert_for_masked_lm_output.hidden_states,
|
111 |
-
bert_for_masked_lm_output.attentions,
|
|
|
112 |
|
113 |
-
def get_projected_text_embeddings(
|
|
|
|
|
114 |
"""
|
115 |
Returns l2-normalised projected cls token embeddings for the given input token ids and attention mask.
|
116 |
The joint latent space is trained using a contrastive objective between image and text data modalities.
|
117 |
|
118 |
:param input_ids: (batch_size, sequence_length)
|
119 |
:param attention_mask: (batch_size, sequence_length)
|
|
|
120 |
:return: (batch_size, projection_size)
|
121 |
"""
|
122 |
|
123 |
-
outputs = self.forward(
|
124 |
-
|
|
|
125 |
assert isinstance(outputs, CXRBertOutput)
|
126 |
|
127 |
-
|
128 |
-
|
|
|
|
|
|
|
|
|
|
|
|
3 |
# Licensed under the MIT License (MIT). See LICENSE in the repo root for license information.
|
4 |
# ------------------------------------------------------------------------------------------
|
5 |
|
6 |
+
from dataclasses import dataclass
|
7 |
from typing import Any, Optional, Tuple, Union
|
8 |
|
9 |
import torch
|
|
|
17 |
|
18 |
BERTTupleOutput = Tuple[T, T, T, T, T]
|
19 |
|
20 |
+
|
21 |
+
@dataclass
|
22 |
class CXRBertOutput(ModelOutput):
|
23 |
last_hidden_state: torch.FloatTensor
|
24 |
+
logits: Optional[torch.FloatTensor] = None
|
25 |
cls_projected_embedding: Optional[torch.FloatTensor] = None
|
26 |
hidden_states: Optional[Tuple[torch.FloatTensor]] = None
|
27 |
attentions: Optional[Tuple[torch.FloatTensor]] = None
|
28 |
|
29 |
|
30 |
class BertProjectionHead(nn.Module):
|
31 |
+
"""Projection head to be used with BERT CLS token.
|
32 |
+
|
33 |
+
This is similar to ``BertPredictionHeadTransform`` in HuggingFace.
|
34 |
+
|
35 |
+
:param config: Configuration for BERT.
|
36 |
+
"""
|
37 |
+
|
38 |
def __init__(self, config: CXRBertConfig) -> None:
|
39 |
super().__init__()
|
40 |
self.dense_to_hidden = nn.Linear(config.hidden_size, config.projection_size)
|
|
|
55 |
"""
|
56 |
Implements the CXR-BERT model outlined in the manuscript:
|
57 |
Boecking et al. "Making the Most of Text Semantics to Improve Biomedical Vision-Language Processing", 2022
|
58 |
+
https://link.springer.com/chapter/10.1007/978-3-031-20059-5_1
|
59 |
|
60 |
+
Extends the HuggingFace BertForMaskedLM model by adding a separate projection head. The projection "[CLS]" token is
|
61 |
+
used to align the latent vectors of image and text modalities.
|
62 |
"""
|
63 |
|
64 |
+
config_class = CXRBertConfig # type: ignore
|
65 |
|
66 |
def __init__(self, config: CXRBertConfig):
|
67 |
super().__init__(config)
|
|
|
83 |
return_dict: Optional[bool] = None,
|
84 |
**kwargs: Any
|
85 |
) -> Union[BERTTupleOutput, CXRBertOutput]:
|
|
|
86 |
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
87 |
|
88 |
+
bert_for_masked_lm_output = super().forward(
|
89 |
+
input_ids=input_ids,
|
90 |
+
attention_mask=attention_mask,
|
91 |
+
token_type_ids=token_type_ids,
|
92 |
+
position_ids=position_ids,
|
93 |
+
head_mask=head_mask,
|
94 |
+
inputs_embeds=inputs_embeds,
|
95 |
+
output_attentions=output_attentions,
|
96 |
+
output_hidden_states=True,
|
97 |
+
return_dict=True,
|
98 |
+
)
|
99 |
|
100 |
last_hidden_state = bert_for_masked_lm_output.hidden_states[-1]
|
101 |
+
cls_projected_embedding = (
|
102 |
+
self.cls_projection_head(last_hidden_state[:, 0, :]) if output_cls_projected_embedding else None
|
103 |
+
)
|
104 |
|
105 |
if return_dict:
|
106 |
return CXRBertOutput(
|
|
|
116 |
bert_for_masked_lm_output.logits,
|
117 |
cls_projected_embedding,
|
118 |
bert_for_masked_lm_output.hidden_states,
|
119 |
+
bert_for_masked_lm_output.attentions,
|
120 |
+
)
|
121 |
|
122 |
+
def get_projected_text_embeddings(
|
123 |
+
self, input_ids: torch.Tensor, attention_mask: torch.Tensor, normalize_embeddings: bool = True
|
124 |
+
) -> torch.Tensor:
|
125 |
"""
|
126 |
Returns l2-normalised projected cls token embeddings for the given input token ids and attention mask.
|
127 |
The joint latent space is trained using a contrastive objective between image and text data modalities.
|
128 |
|
129 |
:param input_ids: (batch_size, sequence_length)
|
130 |
:param attention_mask: (batch_size, sequence_length)
|
131 |
+
:param normalize_embeddings: Whether to l2-normalise the embeddings.
|
132 |
:return: (batch_size, projection_size)
|
133 |
"""
|
134 |
|
135 |
+
outputs = self.forward(
|
136 |
+
input_ids=input_ids, attention_mask=attention_mask, output_cls_projected_embedding=True, return_dict=True
|
137 |
+
)
|
138 |
assert isinstance(outputs, CXRBertOutput)
|
139 |
|
140 |
+
cls_projected_embedding = outputs.cls_projected_embedding
|
141 |
+
assert cls_projected_embedding is not None
|
142 |
+
|
143 |
+
if normalize_embeddings:
|
144 |
+
return F.normalize(cls_projected_embedding, dim=1)
|
145 |
+
|
146 |
+
return cls_projected_embedding
|