File size: 3,927 Bytes
57bafce
 
 
bd0c703
 
57bafce
582b2f2
cf62d27
57bafce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd0c703
57bafce
 
 
 
 
bd0c703
57bafce
bd0c703
 
 
57bafce
 
 
 
bd0c703
57bafce
 
 
 
 
bd0c703
 
 
 
 
 
 
 
 
 
0bd9ff0
dd9c120
bd0c703
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0bd9ff0
dd9c120
bd0c703
 
 
 
 
 
 
 
 
 
 
 
 
 
 
57bafce
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
from transformers import RobertaTokenizer, RobertaForSequenceClassification, RobertaModel
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.utils.data import TensorDataset, DataLoader


device = torch.device("cpu")
class MLP(nn.Module):
    def __init__(self, input_dim):
        super(MLP, self).__init__()
        self.fc1 = nn.Linear(input_dim, 256)
        self.fc2 = nn.Linear(256, 2)
        self.gelu = nn.GELU()

    def forward(self, x):
        x = self.gelu(self.fc1(x))
        x = self.fc2(x)
        return x
def extract_features(text):

    tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
    model = RobertaModel.from_pretrained("roberta-base").to(device)
    tokenized_text = tokenizer.encode(text, truncation=True, max_length=512, return_tensors="pt")
    outputs = model(tokenized_text)
    last_hidden_states = outputs.last_hidden_state
    TClassification = last_hidden_states[:, 0, :].squeeze().detach().numpy()
    return TClassification

def RobertaSentinelOpenGPTInference(input_text):
    features = extract_features(input_text)
    loaded_model = MLP(768).to(device)
    loaded_model.load_state_dict(torch.load("SentinelCheckpoint/RobertaSentinelOpenGPT.pth", map_location=device))

    # Define the tokenizer and model for feature extraction
    with torch.no_grad():
        inputs = torch.tensor(features).to(device)
        outputs = loaded_model(inputs.float())
        _, predicted = torch.max(outputs, 0)

        Probs = (F.softmax(outputs, dim=0).cpu().numpy())

    return Probs

def RobertaSentinelCSAbstractInference(input_text):
    features = extract_features(input_text)
    loaded_model = MLP(768).to(device)
    loaded_model.load_state_dict(torch.load("SentinelCheckpoint/RobertaSentinelCSAbstract.pth", map_location=device))

    # Define the tokenizer and model for feature extraction
    with torch.no_grad():
        inputs = torch.tensor(features).to(device)
        outputs = loaded_model(inputs.float())
        _, predicted = torch.max(outputs, 0)

        Probs = (F.softmax(outputs, dim=0).cpu().numpy())

    return Probs


def RobertaClassifierOpenGPTInference(input_text):
    tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
    model_path = "ClassifierCheckpoint/RobertaClassifierOpenGPT.pth"
    model = RobertaForSequenceClassification.from_pretrained('roberta-base', num_labels=2)
    model.load_state_dict(torch.load(model_path), map_location=torch.device("cpu"))
    model = model.to(torch.device('cpu'))
    model.eval()


    tokenized_input = tokenizer(input_text, truncation=True, padding=True, max_length=512, return_tensors='pt')
    input_ids = tokenized_input['input_ids'].to(torch.device('cpu'))
    attention_mask = tokenized_input['attention_mask'].to(torch.device('cpu'))

    # Make a prediction
    with torch.no_grad():
        outputs = model(input_ids, attention_mask=attention_mask)
    logits = outputs.logits
    Probs = F.softmax(logits, dim=1).cpu().numpy()[0]

    return Probs


def RobertaClassifierCSAbstractInference(input_text):
    tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
    model_path = "ClassifierCheckpoint/RobertaClassifierCSAbstract.pth"
    model = RobertaForSequenceClassification.from_pretrained('roberta-base', num_labels=2)
    model.load_state_dict(torch.load(model_path), map_location=torch.device("cpu"))
    model = model.to(torch.device('cpu'))
    model.eval()


    tokenized_input = tokenizer(input_text, truncation=True, padding=True, max_length=512, return_tensors='pt')
    input_ids = tokenized_input['input_ids'].to(torch.device('cpu'))
    attention_mask = tokenized_input['attention_mask'].to(torch.device('cpu'))

    # Make a prediction
    with torch.no_grad():
        outputs = model(input_ids, attention_mask=attention_mask)
    logits = outputs.logits
    Probs = F.softmax(logits, dim=1).cpu().numpy()[0]

    return Probs