Updated ganbert.py
Browse files- ganbert.py +10 -24
ganbert.py
CHANGED
@@ -30,6 +30,7 @@ from transformers import (
|
|
30 |
set_seed,
|
31 |
get_constant_schedule_with_warmup,
|
32 |
Trainer,TrainingArguments,EarlyStoppingCallback)
|
|
|
33 |
from datasets import Dataset
|
34 |
import torch.nn as nn
|
35 |
import torch.nn.functional as F
|
@@ -77,29 +78,14 @@ class GAN(PreTrainedModel):
|
|
77 |
self.generator.cuda()
|
78 |
self.discriminator.cuda()
|
79 |
self.transformer.cuda()
|
80 |
-
def forward(self
|
81 |
-
input_ids: Optional[torch.Tensor] = None,
|
82 |
-
attention_mask: Optional[torch.Tensor] = None,
|
83 |
-
token_type_ids: Optional[torch.Tensor] = None,
|
84 |
-
position_ids: Optional[torch.Tensor] = None,
|
85 |
-
head_mask: Optional[torch.Tensor] = None,
|
86 |
-
inputs_embeds: Optional[torch.Tensor] = None,
|
87 |
-
encoder_hidden_states: Optional[torch.Tensor] = None,
|
88 |
-
encoder_attention_mask: Optional[torch.Tensor] = None,
|
89 |
-
past_key_values: Optional[List[torch.FloatTensor]] = None,
|
90 |
-
use_cache: Optional[bool] = None,
|
91 |
-
output_attentions: Optional[bool] = None,
|
92 |
-
output_hidden_states: Optional[bool] = None,
|
93 |
-
return_dict: Optional[bool] = None,
|
94 |
-
):
|
95 |
# Encode real data in the Transformer
|
96 |
-
real_batch_size = input_ids.shape[0]
|
97 |
-
model_outputs = self.transformer(
|
98 |
# print('got transformer output')
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
return model_outputs[0]
|
|
|
30 |
set_seed,
|
31 |
get_constant_schedule_with_warmup,
|
32 |
Trainer,TrainingArguments,EarlyStoppingCallback)
|
33 |
+
|
34 |
from datasets import Dataset
|
35 |
import torch.nn as nn
|
36 |
import torch.nn.functional as F
|
|
|
78 |
self.generator.cuda()
|
79 |
self.discriminator.cuda()
|
80 |
self.transformer.cuda()
|
81 |
+
def forward(self,**kwargs):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
82 |
# Encode real data in the Transformer
|
83 |
+
# real_batch_size = input_ids.shape[0]
|
84 |
+
model_outputs = self.transformer(**kwargs)
|
85 |
# print('got transformer output')
|
86 |
+
# hidden_states = torch.mean(model_outputs[0],dim=1)
|
87 |
+
# noise = torch.zeros(real_batch_size, self.ns, device=self.dv).uniform_(0, 1).to(self.dv)
|
88 |
+
# gen_rep = self.generator(noise)
|
89 |
+
# disciminator_input = torch.cat([hidden_states, gen_rep], dim=0)
|
90 |
+
# features, logits, probs = self.discriminator(disciminator_input)
|
91 |
+
return model_outputs
|
|