Create app.py

app.py

@@ -0,0 +1,131 @@
+import gradio as gr
+import torch
+from transformers import AutoTokenizer, AutoModel
+import torch.nn as nn
+import numpy as np
+
+class HybridCodeClassifier(nn.Module):
+    def __init__(self, model_name="microsoft/codebert-base", num_labels=4):
+        super().__init__()
+        self.encoder = AutoModel.from_pretrained(model_name)
+        self.classifier = nn.Linear(768, num_labels)
+        
+    def forward(self, input_ids, attention_mask):
+        outputs = self.encoder(input_ids=input_ids, attention_mask=attention_mask)
+        return self.classifier(outputs.last_hidden_state[:, 0, :])
+
+# Label mappings for Task C
+label_mappings = {
+    0: "Human-written 👨‍💻",
+    1: "Machine-generated 🤖", 
+    2: "Hybrid 🔄",
+    3: "Adversarial ⚔️"
+}
+
+label_descriptions = {
+    0: "• Natural code patterns\n• Imperfections and TODOs\n• Personal coding style\n• Practical solutions",
+    1: "• Perfect structure\n• Comprehensive docs\n• Consistent formatting\n• Over-engineered",
+    2: "• Mixed patterns\n• Some AI elements\n• Some human elements\n• Inconsistent style", 
+    3: "• Designed to mimic humans\n• Strategic imperfections\n• Hard to detect\n• Evasive patterns"
+}
+
+@torch.no_grad()
+def load_taskC_model():
+    model = HybridCodeClassifier(num_labels=4)
+    
+    try:
+        from huggingface_hub import hf_hub_download
+        model_path = hf_hub_download(
+            repo_id="KrishnaKarthik/ai-code-detector",
+            filename="taskC_model.pth"
+        )
+        model.load_state_dict(torch.load(model_path, map_location="cpu"))
+        print("✅ Loaded Task C Hybrid Code Detector!")
+    except Exception as e:
+        print(f"❌ Error: {str(e)}")
+        return None
+    
+    model.eval()
+    return model
+
+model_taskC = load_taskC_model()
+tokenizer_taskC = AutoTokenizer.from_pretrained("microsoft/codebert-base")
+
+def detect_hybrid_code(code):
+    """Detect human, machine, hybrid, or adversarial code"""
+    if not code.strip():
+        return "Please enter code", "", "", ""
+    
+    try:
+        inputs = tokenizer_taskC(code, return_tensors="pt", truncation=True, max_length=512)
+        
+        with torch.no_grad():
+            outputs = model_taskC(**inputs)
+            probabilities = torch.softmax(outputs, dim=1)
+            probs = probabilities[0].numpy()
+        
+        # Get all predictions
+        results = "🔍 DETECTION RESULTS:\n"
+        results += "=" * 50 + "\n"
+        
+        for i, (label_id, label_name) in enumerate(label_mappings.items()):
+            prob = probs[label_id]
+            results += f"{i+1}. {label_name:20} {prob:.1%}\n"
+        
+        # Main prediction
+        main_pred_idx = np.argmax(probs)
+        main_pred_label = label_mappings[main_pred_idx]
+        main_pred_prob = probs[main_pred_idx]
+        main_description = label_descriptions[main_pred_idx]
+        
+        # Confidence
+        if main_pred_prob >= 0.8:
+            confidence = "🟢 HIGH confidence"
+        elif main_pred_prob >= 0.6:
+            confidence = "🟡 MEDIUM confidence"
+        else:
+            confidence = "🔴 LOW confidence"
+        
+        return results, main_pred_label, main_description, confidence
+        
+    except Exception as e:
+        return f"Error: {str(e)}", "Error", "", ""
+
+# Gradio Interface
+with gr.Blocks(title="Hybrid Code Detector", theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🔄 Hybrid Code Detector")
+    gr.Markdown("Classify code as: **Human** 👨‍💻 | **Machine** 🤖 | **Hybrid** 🔄 | **Adversarial** ⚔️")
+    
+    with gr.Row():
+        code_input = gr.Textbox(
+            label="Paste code to analyze",
+            placeholder="def hello_world():\n    print('Hello, World!')",
+            lines=10
+        )
+    
+    analyze_btn = gr.Button("Analyze Code", variant="primary", size="lg")
+    
+    with gr.Row():
+        with gr.Column():
+            results_output = gr.Textbox(label="Detection Results", lines=8)
+            confidence_output = gr.Textbox(label="Confidence Level")
+        with gr.Column():
+            prediction_output = gr.Textbox(label="Primary Prediction")
+            description_output = gr.Textbox(label="Characteristics", lines=4)
+    
+    gr.Markdown("### 💡 Examples to Test:")
+    examples = [
+        ["def calc(x):\n    # quick hack\n    result = x * 2\n    if x > 10:\n        result += 5\n    return result", "Human-like"],
+        ["def calculate_sum(numbers):\n    '''\n    Calculate the sum of all numbers in the input list.\n    \n    Args:\n        numbers (List[int]): Input list of numbers\n    \n    Returns:\n        int: Sum of all numbers\n    '''\n    return sum(numbers)", "AI-like"],
+    ]
+    
+    gr.Examples(examples=examples, inputs=code_input)
+    
+    analyze_btn.click(
+        fn=detect_hybrid_code,
+        inputs=code_input,
+        outputs=[results_output, prediction_output, description_output, confidence_output]
+    )
+
+if __name__ == "__main__":
+    demo.launch()