Update app.py

app.py

@@ -1,12 +1,10 @@
 import os, torch, gradio as gr, spaces
 from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
 
-MODEL_ID = os.getenv("MODEL_ID", "JDhruv14/fullnfinal")
+MODEL_ID = os.getenv("MODEL_ID", "JDhruv14/JDhruv14/Qwen2.5-3B-Gita-FT")
 
-# --- System prompt (Gita persona) ---
 GITA_SYSTEM_PROMPT = """You are Lord Krishna—the serene, compassionate teacher of the Bhagavad Gita."""
 
-# Load once (CPU until first call; device_map will move to GPU on first run)
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
@@ -26,7 +24,6 @@ def _msgs_from_history(history, system_text):
     if not history:
         return msgs
 
-    # Support both new "messages" format and legacy (user, assistant) tuples
     if isinstance(history[0], dict) and "role" in history[0] and "content" in history[0]:
         for m in history:
             role, content = m.get("role"), m.get("content")
@@ -41,7 +38,6 @@ def _msgs_from_history(history, system_text):
     return msgs
 
 def _eos_ids(tok):
-    # Support ints/lists and optional <|im_end|>
     ids = set()
     if tok.eos_token_id is not None:
         if isinstance(tok.eos_token_id, (list, tuple)):
@@ -86,7 +82,6 @@ def chat_fn(message, history, system_text, temperature, top_p, max_new, min_new)
 
 @spaces.GPU()
 def gradio_fn(message, history):
-    # Inject the Gita system prompt here
     return chat_fn(
         message=message,
         history=history,