Create app.py

app.py

@@ -0,0 +1,91 @@
+# ==============================================================================
+# V5 GRADIO DEPLOYMENT SCRIPT
+# ==============================================================================
+# This script creates a web UI to test your v5 model.
+#
+# TO DEPLOY ON HUGGING FACE SPACES:
+# 1. Create a new Space and choose the "Gradio" SDK.
+# 2. Select the free "T4 small" GPU hardware.
+# 3. Create a file named `app.py` and paste this code into it.
+# 4. Create a `requirements.txt` file and add the libraries listed below.
+# ==============================================================================
+
+# requirements.txt file contents:
+# gradio
+# transformers
+# peft
+# accelerate
+# bitsandbytes
+# torch
+
+import torch
+import gradio as gr
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+
+# --- Configuration ---
+# Your final, v5 model on the Hugging Face Hub
+MODEL_ID = "Arko007/my-awesome-code-assistant-v5" 
+BASE_MODEL_ID = "codellama/CodeLlama-7b-hf"
+
+# --- Load the Model (Memory-Optimized) ---
+print("Setting up 4-bit quantization...")
+quantization_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.float16,
+)
+
+print(f"Loading fine-tuned model: {MODEL_ID}...")
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    quantization_config=quantization_config,
+    device_map="auto"
+)
+
+tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL_ID)
+tokenizer.pad_token = tokenizer.eos_token
+
+print("✅ Model loaded successfully!")
+
+# --- Inference Function ---
+def generate_code(instruction):
+    """
+    Generates code from an instruction using the v5 model.
+    """
+    prompt = f"""### Instruction:
+{instruction}
+
+### Response:"""
+    
+    inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
+    
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=1024, # Give it plenty of room to write
+        temperature=0.1,
+        top_p=0.9,
+        eos_token_id=tokenizer.eos_token_id
+    )
+    
+    response_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    # Extract only the code part of the response
+    code_part = response_text.split("### Response:")[1].strip()
+    return code_part
+
+# --- Create and Launch the Gradio Web App ---
+print("Launching Gradio app...")
+
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🤖 My Awesome Code Assistant (v5)")
+    gr.Markdown("Enter an instruction and I'll generate the code for you!")
+    
+    with gr.Row():
+        instruction_box = gr.Textbox(lines=5, label="Instruction", placeholder="e.g., Write a Python function to sort a list of numbers.")
+        output_box = gr.Code(label="Generated Code", language="python")
+        
+    generate_button = gr.Button("Generate Code", variant="primary")
+    
+    generate_button.click(fn=generate_code, inputs=instruction_box, outputs=output_box)
+
+# This will launch the app when deployed on Hugging Face Spaces
+demo.launch()