Spaces:
Running
Running
ThanaritKanjanametawat
commited on
Commit
•
1bb143a
1
Parent(s):
1fb6014
Move everything to CPU
Browse files- ModelDriver.py +10 -9
- Test.py +1 -1
ModelDriver.py
CHANGED
@@ -5,7 +5,8 @@ import torch.nn.functional as F
|
|
5 |
from torch.utils.data import TensorDataset, DataLoader
|
6 |
|
7 |
|
8 |
-
device = torch.device("cpu")
|
|
|
9 |
class MLP(nn.Module):
|
10 |
def __init__(self, input_dim):
|
11 |
super(MLP, self).__init__()
|
@@ -62,14 +63,14 @@ def RobertaClassifierOpenGPTInference(input_text):
|
|
62 |
tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
|
63 |
model_path = "ClassifierCheckpoint/RobertaClassifierOpenGPT.pth"
|
64 |
model = RobertaForSequenceClassification.from_pretrained('roberta-base', num_labels=2)
|
65 |
-
model.load_state_dict(torch.load(model_path, map_location=
|
66 |
-
model = model.to(
|
67 |
model.eval()
|
68 |
|
69 |
|
70 |
tokenized_input = tokenizer(input_text, truncation=True, padding=True, max_length=512, return_tensors='pt')
|
71 |
-
input_ids = tokenized_input['input_ids'].to(
|
72 |
-
attention_mask = tokenized_input['attention_mask'].to(
|
73 |
|
74 |
# Make a prediction
|
75 |
with torch.no_grad():
|
@@ -84,14 +85,14 @@ def RobertaClassifierCSAbstractInference(input_text):
|
|
84 |
tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
|
85 |
model_path = "ClassifierCheckpoint/RobertaClassifierCSAbstract.pth"
|
86 |
model = RobertaForSequenceClassification.from_pretrained('roberta-base', num_labels=2)
|
87 |
-
model.load_state_dict(torch.load(model_path, map_location=
|
88 |
-
model = model.to(
|
89 |
model.eval()
|
90 |
|
91 |
|
92 |
tokenized_input = tokenizer(input_text, truncation=True, padding=True, max_length=512, return_tensors='pt')
|
93 |
-
input_ids = tokenized_input['input_ids'].to(
|
94 |
-
attention_mask = tokenized_input['attention_mask'].to(
|
95 |
|
96 |
# Make a prediction
|
97 |
with torch.no_grad():
|
|
|
5 |
from torch.utils.data import TensorDataset, DataLoader
|
6 |
|
7 |
|
8 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
9 |
+
# device = torch.device("cpu")
|
10 |
class MLP(nn.Module):
|
11 |
def __init__(self, input_dim):
|
12 |
super(MLP, self).__init__()
|
|
|
63 |
tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
|
64 |
model_path = "ClassifierCheckpoint/RobertaClassifierOpenGPT.pth"
|
65 |
model = RobertaForSequenceClassification.from_pretrained('roberta-base', num_labels=2)
|
66 |
+
model.load_state_dict(torch.load(model_path, map_location=device))
|
67 |
+
model = model.to(device)
|
68 |
model.eval()
|
69 |
|
70 |
|
71 |
tokenized_input = tokenizer(input_text, truncation=True, padding=True, max_length=512, return_tensors='pt')
|
72 |
+
input_ids = tokenized_input['input_ids'].to(device)
|
73 |
+
attention_mask = tokenized_input['attention_mask'].to(device)
|
74 |
|
75 |
# Make a prediction
|
76 |
with torch.no_grad():
|
|
|
85 |
tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
|
86 |
model_path = "ClassifierCheckpoint/RobertaClassifierCSAbstract.pth"
|
87 |
model = RobertaForSequenceClassification.from_pretrained('roberta-base', num_labels=2)
|
88 |
+
model.load_state_dict(torch.load(model_path, map_location=device))
|
89 |
+
model = model.to(device)
|
90 |
model.eval()
|
91 |
|
92 |
|
93 |
tokenized_input = tokenizer(input_text, truncation=True, padding=True, max_length=512, return_tensors='pt')
|
94 |
+
input_ids = tokenized_input['input_ids'].to(device)
|
95 |
+
attention_mask = tokenized_input['attention_mask'].to(device)
|
96 |
|
97 |
# Make a prediction
|
98 |
with torch.no_grad():
|
Test.py
CHANGED
@@ -20,7 +20,7 @@ Input_Text = "I want to do this data"
|
|
20 |
# print(f"Confidence:", max(Probs))
|
21 |
|
22 |
print("RobertaClassifierCSAbstractInference")
|
23 |
-
Probs =
|
24 |
Pred = "Human Written" if not np.argmax(Probs) else "Machine Generated"
|
25 |
|
26 |
print(Probs)
|
|
|
20 |
# print(f"Confidence:", max(Probs))
|
21 |
|
22 |
print("RobertaClassifierCSAbstractInference")
|
23 |
+
Probs = RobertaClassifierCSAbstractInference(Input_Text)
|
24 |
Pred = "Human Written" if not np.argmax(Probs) else "Machine Generated"
|
25 |
|
26 |
print(Probs)
|