update app by using token

app.py

@@ -15,8 +15,10 @@ if not USE_REMOTE_OLLAMA:
     # Transformers fallback for Spaces (CPU-friendly small instruct model)
     from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
     HF_CHAT_MODEL = os.getenv("HF_CHAT_MODEL", "google/gemma-2-2b-it")  # small instruct model that runs on CPU
-    _tok = AutoTokenizer.from_pretrained(HF_CHAT_MODEL)
-    _mdl = AutoModelForCausalLM.from_pretrained(HF_CHAT_MODEL, torch_dtype="auto", device_map="auto")
+    HF_TOKEN = os.getenv("HF_TOKEN")
+    
+    _tok = AutoTokenizer.from_pretrained(HF_CHAT_MODEL, token=HF_TOKEN)
+    _mdl = AutoModelForCausalLM.from_pretrained(HF_CHAT_MODEL, token=HF_TOKEN, torch_dtype="auto", device_map="auto")
     gen = pipeline("text-generation", model=_mdl, tokenizer=_tok, max_new_tokens=256)