1

app.py

@@ -1,118 +0,0 @@
-import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
-from peft import PeftModel
-import torch
-import os
-
-# Constants
-MODEL_NAME = "Salesforce/codegen-350M-mono"
-LORA_PATH = "fine-tuned-model"
-
-# Initialize tokenizer and model
-print("Loading tokenizer...")
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, trust_remote_code=True)
-
-print("Loading base model...")
-base_model = AutoModelForCausalLM.from_pretrained(
-    MODEL_NAME,
-    trust_remote_code=True,
-    device_map="auto",
-    torch_dtype=torch.float16
-)
-
-print("Loading fine-tuned model...")
-model = PeftModel.from_pretrained(
-    base_model,
-    LORA_PATH,
-    device_map="auto",
-    torch_dtype=torch.float16
-)
-
-def generate_response(task_description, code_snippet, request_type, mode="concise"):
-    try:
-        # Format the prompt based on request type
-        if request_type == "hint":
-            prompt = f"Task: {task_description}\nCode:\n{code_snippet}\nHINT:"
-        elif request_type == "feedback":
-            prompt = f"Task: {task_description}\nCode:\n{code_snippet}\nFEEDBACK:"
-        else:  # follow-up
-            prompt = f"Task: {task_description}\nCode:\n{code_snippet}\nFOLLOW-UP:"
-        
-        # Encode and generate
-        inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-        
-        outputs = model.generate(
-            **inputs,
-            max_new_tokens=256 if mode == "detailed" else 128,
-            do_sample=True,
-            temperature=0.7,
-            top_p=0.95,
-        )
-        
-        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-        
-        # Extract the relevant part of the response
-        if request_type == "hint" and "HINT:" in response:
-            response = response.split("HINT:", 1)[1].strip()
-        elif request_type == "feedback" and "FEEDBACK:" in response:
-            response = response.split("FEEDBACK:", 1)[1].strip()
-        elif request_type == "follow-up" and "FOLLOW-UP:" in response:
-            response = response.split("FOLLOW-UP:", 1)[1].strip()
-        
-        return response
-    except Exception as e:
-        return f"An error occurred: {str(e)}"
-
-# Create Gradio interface
-with gr.Blocks(title="Live Coding HR Assistant") as demo:
-    gr.Markdown("# 💻 Live Coding HR Assistant")
-    gr.Markdown("Get hints, feedback, and follow-up questions for your coding tasks!")
-    
-    with gr.Row():
-        with gr.Column():
-            task_description = gr.Textbox(
-                label="Task Description",
-                value="",
-                lines=5
-            )
-            code_snippet = gr.Code(
-                label="Code Snippet",
-                language="python",
-                value=""
-            )
-            request_type = gr.Radio(
-                choices=["hint", "feedback", "follow-up"],
-                label="What would you like?",
-                value="hint"
-            )
-            mode = gr.Radio(
-                choices=["concise", "detailed"],
-                label="Response Style",
-                value="concise"
-            )
-            submit_btn = gr.Button("Get Response", variant="primary")
-        
-        with gr.Column():
-            output = gr.Textbox(
-                label="AI Response",
-                lines=8,
-                value=""
-            )
-    
-    submit_btn.click(
-        fn=generate_response,
-        inputs=[task_description, code_snippet, request_type, mode],
-        outputs=output,
-        api_name="predict",
-        concurrency_limit=1  # Set concurrency limit here
-    )
-
-# Configure queue and launch
-demo.queue(
-    max_size=10  # Removed concurrency_count
-).launch(
-    server_name="0.0.0.0",
-    server_port=7860,
-    share=True
-    # You can optionally add max_threads here if needed, e.g., max_threads=10
-)