initial upload
Browse files- config.json +2 -1
- configuration_ltgbert.py +2 -0
- modeling_ltgbert.py +2 -1
- pytorch_model.bin +2 -2
- spacial_tokens_map.json +1 -1
- tokenizer.json +0 -0
- tokenizer_config.json +7 -7
config.json
CHANGED
@@ -22,5 +22,6 @@
|
|
22 |
"num_hidden_layers": 12,
|
23 |
"position_bucket_size": 32,
|
24 |
"torch_dtype": "float32",
|
25 |
-
"vocab_size":
|
|
|
26 |
}
|
|
|
22 |
"num_hidden_layers": 12,
|
23 |
"position_bucket_size": 32,
|
24 |
"torch_dtype": "float32",
|
25 |
+
"vocab_size": 16384,
|
26 |
+
"temperature": 2.5
|
27 |
}
|
configuration_ltgbert.py
CHANGED
@@ -17,6 +17,7 @@ class LtgbertConfig(PretrainedConfig):
|
|
17 |
num_hidden_layers=12,
|
18 |
layer_norm_eps=1.0e-7,
|
19 |
output_all_encoded_layers=True,
|
|
|
20 |
**kwargs,
|
21 |
):
|
22 |
super().__init__(**kwargs)
|
@@ -32,3 +33,4 @@ class LtgbertConfig(PretrainedConfig):
|
|
32 |
self.output_all_encoded_layers = output_all_encoded_layers
|
33 |
self.position_bucket_size = position_bucket_size
|
34 |
self.layer_norm_eps = layer_norm_eps
|
|
|
|
17 |
num_hidden_layers=12,
|
18 |
layer_norm_eps=1.0e-7,
|
19 |
output_all_encoded_layers=True,
|
20 |
+
temperature=1.0,
|
21 |
**kwargs,
|
22 |
):
|
23 |
super().__init__(**kwargs)
|
|
|
33 |
self.output_all_encoded_layers = output_all_encoded_layers
|
34 |
self.position_bucket_size = position_bucket_size
|
35 |
self.layer_norm_eps = layer_norm_eps
|
36 |
+
self.temperature = temperature
|
modeling_ltgbert.py
CHANGED
@@ -443,6 +443,7 @@ class Classifier(nn.Module):
|
|
443 |
def __init__(self, config, num_labels: int):
|
444 |
super().__init__()
|
445 |
|
|
|
446 |
drop_out = getattr(config, "cls_dropout", None)
|
447 |
drop_out = config.hidden_dropout_prob if drop_out is None else drop_out
|
448 |
|
@@ -456,7 +457,7 @@ class Classifier(nn.Module):
|
|
456 |
)
|
457 |
|
458 |
def forward(self, x):
|
459 |
-
x = self.nonlinearity(x)
|
460 |
return x
|
461 |
|
462 |
|
|
|
443 |
def __init__(self, config, num_labels: int):
|
444 |
super().__init__()
|
445 |
|
446 |
+
self.temperature = config.temperature
|
447 |
drop_out = getattr(config, "cls_dropout", None)
|
448 |
drop_out = config.hidden_dropout_prob if drop_out is None else drop_out
|
449 |
|
|
|
457 |
)
|
458 |
|
459 |
def forward(self, x):
|
460 |
+
x = self.nonlinearity(x) / self.temperature
|
461 |
return x
|
462 |
|
463 |
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e5aaa5fba47a98d324a6af79ec8a03fe32c6abf1a483c0274075d5faa9e5d6a6
|
3 |
+
size 503044925
|
spacial_tokens_map.json
CHANGED
@@ -1 +1 @@
|
|
1 |
-
{"bos_token": "
|
|
|
1 |
+
{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "sep_token": "</s>", "pad_token": "<oad>", "cls_token": "<s>", "mask_token": "<mask>"}
|
tokenizer.json
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
tokenizer_config.json
CHANGED
@@ -1,10 +1,10 @@
|
|
1 |
{
|
2 |
"tokenizer_class": "PreTrainedTokenizerFast",
|
3 |
-
"bos_token": "
|
4 |
-
"eos_token": "
|
5 |
-
"unk_token": "
|
6 |
-
"sep_token": "
|
7 |
-
"pad_token": "
|
8 |
-
"cls_token": "
|
9 |
-
"mask_token": "
|
10 |
}
|
|
|
1 |
{
|
2 |
"tokenizer_class": "PreTrainedTokenizerFast",
|
3 |
+
"bos_token": "<s>",
|
4 |
+
"eos_token": "</s>",
|
5 |
+
"unk_token": "<unk>",
|
6 |
+
"sep_token": "</s>",
|
7 |
+
"pad_token": "<pad>",
|
8 |
+
"cls_token": "<s>",
|
9 |
+
"mask_token": "<mask>"
|
10 |
}
|