Update app.py

app.py

@@ -1,39 +1,39 @@
 import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+from peft import PeftModel
 from threading import Thread
 
-# 1. Point the script directly to your uploaded Hugging Face model
-MODEL_ID = "Cydercoder/qwen2.5-coder-3b"
+# 1. Map both coordinates: The base model engine and your custom adapter layer
+BASE_MODEL = "Qwen/Qwen2.5-Coder-3B-Instruct"
+ADAPTER_MODEL = "Cydercoder/qwen2.5-coder-3b"
 
-print("Loading tokenizer...")
-tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
+print("Loading official base tokenizer...")
+tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
 
-print("Loading model on CPU via standard clean mapping...")
-# Forcing float32 without low_cpu_mem_usage prevents the Transformers v5 thread crash
-model = AutoModelForCausalLM.from_pretrained(
-    MODEL_ID,
+print("Loading public base model on CPU...")
+base_model = AutoModelForCausalLM.from_pretrained(
+    BASE_MODEL,
     torch_dtype=torch.float32,
     device_map="cpu"
 )
 
+print("Merging your custom fine-tuned engineering weights...")
+# This layers your specialized tasks right over the active model architecture
+model = PeftModel.from_pretrained(base_model, ADAPTER_MODEL)
+
 def chat_function(message, history):
-    # Construct formatting conversation list matrices
     messages = [
-        {"role": "system", "content": "You are an expert full-stack developer assistant."}
+        {"role": "system", "content": "You are an expert full-stack developer assistant fine-tuned for frontend, backend, animations, and debugging."}
     ]
     
-    # Re-insert existing browser chat history logs
     for user_msg, bot_msg in history:
         messages.append({"role": "user", "content": user_msg})
         messages.append({"role": "assistant", "content": bot_msg})
         
     messages.append({"role": "user", "content": message})
     
-    # Process tokens safely
     inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")
-    
-    # Set up a dynamic background streamer so answers appear word-by-word in browser
     streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
     
     generation_kwargs = dict(
@@ -43,7 +43,6 @@ def chat_function(message, history):
         temperature=0.6,
     )
     
-    # Run text generation in a separate background processor thread
     thread = Thread(target=model.generate, kwargs=generation_kwargs)
     thread.start()
     
@@ -52,7 +51,6 @@ def chat_function(message, history):
         partial_text += new_text
         yield partial_text
 
-# 4. Initialize the native Gradio browser interface
 demo = gr.ChatInterface(
     fn=chat_function,
     title="🤖 Cydercoder Qwen 3B AI Chatbot",