Update app.py

app.py

@@ -1,64 +1,166 @@
+# app.py
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
 import gradio as gr
-from huggingface_hub import InferenceClient
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+# Initialize variables
+model = None
+tokenizer = None
+device = None
 
+# Define function to load model
+def load_model():
+    global model, tokenizer, device
+    
+    # Use GPU if available
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    print(f"Using device: {device}")
+    
+    # Load the Phi-2 model
+    model_id = "microsoft/phi-2"
+    
+    print("Loading Phi-2 model and tokenizer...")
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    model = AutoModelForCausalLM.from_pretrained(
+        model_id, 
+        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+        device_map="auto"  # Better device management for Spaces
+    )
+    print("Model loaded successfully!")
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
+# Define inference function
+def generate_text(prompt, task_type, max_length=300):
+    global model, tokenizer, device
+    
+    # If model hasn't been loaded yet, load it
+    if model is None:
+        load_model()
+    
+    # Set temperature based on task type
+    temperature_map = {
+        "Math Problem": 0.2,
+        "Science Theory": 0.4,
+        "Coding Question": 0.3, 
+        "Reasoning": 0.5,
+        "Creative Writing": 0.8
+    }
+    temperature = temperature_map.get(task_type, 0.5)
+    
+    # Enhance the prompt to request step-by-step solutions
+    enhanced_prompt = f"{prompt}\n\nPlease provide a detailed step-by-step solution with clear reasoning."
+    
+    # Progress update for UI
+    yield "Generating solution..."
+    
+    # Tokenize input
+    inputs = tokenizer(enhanced_prompt, return_tensors="pt").to(device)
+    
+    # Generate output
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs, 
+            max_new_tokens=max_length,
+            temperature=temperature,
+            do_sample=True
+        )
+    
+    # Decode response
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    
+    # If the response doesn't seem to include steps, add formatting for clarity
+    if "step" not in response.lower() and len(response) > 100:
+        # Split into paragraphs and format as steps
+        paragraphs = [p for p in response.split('\n') if p.strip()]
+        formatted_response = ""
+        
+        for i, para in enumerate(paragraphs):
+            if i == 0 and para == enhanced_prompt:
+                continue
+            formatted_response += f"Step {i+1}: {para}\n\n"
+        
+        yield formatted_response
+    else:
         yield response
 
+# Create Gradio interface
+with gr.Blocks(title="Phi-2 Step-by-Step Solution Generator", theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🧠 Phi-2 Step-by-Step Solution Generator")
+    gr.Markdown("""
+    Enter a prompt below and get detailed step-by-step solutions using Microsoft's Phi-2 model.
+    Select the appropriate task type to optimize the model's response.
+    """)
+    
+    with gr.Row():
+        with gr.Column(scale=2):
+            prompt_input = gr.Textbox(
+                label="Prompt",
+                placeholder="Enter your question or problem here...",
+                lines=5
+            )
+            
+            with gr.Row():
+                task_type = gr.Radio(
+                    ["Math Problem", "Science Theory", "Coding Question", "Reasoning", "Creative Writing"],
+                    label="Task Type (sets optimal temperature)",
+                    value="Reasoning"
+                )
+                
+                max_length_slider = gr.Slider(
+                    minimum=100,
+                    maximum=1000,
+                    value=300,
+                    step=50,
+                    label="Maximum Output Length"
+                )
+            
+            with gr.Row():
+                generate_button = gr.Button(
+                    "✨ Generate Step-by-Step Solution", 
+                    variant="primary",
+                    size="lg"
+                )
+                clear_button = gr.Button("Clear", variant="secondary")
+        
+        with gr.Column(scale=3):
+            output_text = gr.Textbox(
+                label="Step-by-Step Solution",
+                lines=15,
+                show_copy_button=True
+            )
+    
+    # Examples with different task types
+    with gr.Accordion("Example Prompts", open=False):
+        gr.Examples(
+            examples=[
+                ["Solve the quadratic equation: 2x² + 5x - 3 = 0", "Math Problem"],
+                ["Explain how photosynthesis works in plants", "Science Theory"],
+                ["Write a function in Python to find the Fibonacci sequence up to n terms", "Coding Question"],
+                ["Why might increasing minimum wage have both positive and negative economic impacts?", "Reasoning"],
+                ["Write a short story about a robot discovering emotions", "Creative Writing"]
+            ],
+            inputs=[prompt_input, task_type]
+        )
+    
+    # Add functionality to buttons
+    generate_button.click(
+        fn=generate_text,
+        inputs=[prompt_input, task_type, max_length_slider],
+        outputs=output_text
+    )
+    
+    # Clear functionality
+    clear_button.click(
+        fn=lambda: ("", "Reasoning"),
+        inputs=[],
+        outputs=[prompt_input, task_type]
+    )
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
-
+    # Adding a note about load times
+    gr.Markdown("""
+    > **Note**: The model loads when you submit your first prompt, which may take 1-2 minutes.
+    > Subsequent generations will be much faster.
+    """)
 
+# Launch the app
 if __name__ == "__main__":
-    demo.launch()
+    demo.queue().launch()