Update Negativ/PythonCode/positive_reframe.py

Negativ/PythonCode/positive_reframe.py

@@ -1,65 +1,65 @@
-import torch
-from transformers import AutoTokenizer, DebertaForSequenceClassification
-
-sentence = "I really love that you suck so bad at this game, you are the worst teammate and i hope you die in a fire"
-
-# Define the model architecture
-num_labels = 2  # Assuming a three-class classification task
-
-# Load the tokenizer
-tokenizer = AutoTokenizer.from_pretrained("microsoft/deberta-base")
-
-# Initialize the model architecture
-model = DebertaForSequenceClassification.from_pretrained("microsoft/deberta-base", num_labels=num_labels)
-
-# Load the saved state dictionary
-state_dict = torch.load('C:/Users/Michael/Documents/GitHub/Negativ/PythonCode/toxic_deberta_tuned.pth')
-
-# Load the state dictionary into the model
-model.load_state_dict(state_dict)
-
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-
-# Ensure the model is set up with the correct number of labels
-assert model.config.num_labels == num_labels, "Model labels mismatch!"
-
-def predict(sentence):
-    # Tokenize the input sentence
-    inputs = tokenizer(sentence, return_tensors="pt", padding='max_length', truncation=True)
-
-    print(inputs)
-
-    inputs = {k: v.to(device) for k, v in inputs.items()}
-    
-    # Move the model to evaluation mode
-    model.eval()
-    
-    # Perform prediction
-    with torch.no_grad():
-        outputs = model(**inputs)
-
-        print(outputs)
-
-        logits = outputs.logits
-        probabilities = torch.softmax(logits, dim=1)
-        
-        print(f"Logits: {logits}")
-        print(f"Probabilities: {probabilities}")
-        
-        highest_logit, prediction = torch.max(logits, dim=-1)
-
-        print(f"Highest Logit: {highest_logit.item()}, Predicted Class: {prediction.item()}")
-    
-    # Interpret the result
-    return "Toxic" if prediction == 1 else "Non-toxic"
-
-def is_toxic(sentence):
-
-    result = predict(sentence)
-
-    print(f"Given Sentence: {sentence}", f"\tPrediction: {result}")
-    
-    if result == "Toxic":
-        return True
-    else:
+import torch
+from transformers import AutoTokenizer, DebertaForSequenceClassification
+
+sentence = "I really love that you suck so bad at this game, you are the worst teammate and i hope you die in a fire"
+
+# Define the model architecture
+num_labels = 2  # Assuming a three-class classification task
+
+# Load the tokenizer
+tokenizer = AutoTokenizer.from_pretrained("microsoft/deberta-base")
+
+# Initialize the model architecture
+model = DebertaForSequenceClassification.from_pretrained("microsoft/deberta-base", num_labels=num_labels)
+
+# Load the saved state dictionary
+state_dict = torch.load('./toxic_deberta_tuned.pth')
+
+# Load the state dictionary into the model
+model.load_state_dict(state_dict)
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+# Ensure the model is set up with the correct number of labels
+assert model.config.num_labels == num_labels, "Model labels mismatch!"
+
+def predict(sentence):
+    # Tokenize the input sentence
+    inputs = tokenizer(sentence, return_tensors="pt", padding='max_length', truncation=True)
+
+    print(inputs)
+
+    inputs = {k: v.to(device) for k, v in inputs.items()}
+    
+    # Move the model to evaluation mode
+    model.eval()
+    
+    # Perform prediction
+    with torch.no_grad():
+        outputs = model(**inputs)
+
+        print(outputs)
+
+        logits = outputs.logits
+        probabilities = torch.softmax(logits, dim=1)
+        
+        print(f"Logits: {logits}")
+        print(f"Probabilities: {probabilities}")
+        
+        highest_logit, prediction = torch.max(logits, dim=-1)
+
+        print(f"Highest Logit: {highest_logit.item()}, Predicted Class: {prediction.item()}")
+    
+    # Interpret the result
+    return "Toxic" if prediction == 1 else "Non-toxic"
+
+def is_toxic(sentence):
+
+    result = predict(sentence)
+
+    print(f"Given Sentence: {sentence}", f"\tPrediction: {result}")
+    
+    if result == "Toxic":
+        return True
+    else:
         return False