Update app.py

app.py

@@ -9,6 +9,7 @@ from huggingface_hub import hf_hub_download
 
 HF_TOKEN = os.environ.get("HF_TOKEN", None)
 MODEL_ID = "google/gemma-2-27b-it"
+REPO_ID = "bartowski/gemma-2-27b-it-GGUF"
 MODEL_NAME = MODEL_ID.split("/")[-1]
 MODEL_FILE = "gemma-2-27b-it-Q4_K_M.gguf"
 
@@ -16,7 +17,7 @@ os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
 
 llm = Llama(
     model_path=hf_hub_download(
-        repo_id=os.environ.get(MODEL_ID),
+        repo_id=os.environ.get(REPO_ID),
         filename=os.environ.get(MODEL_FILE),
     ),
     n_ctx=4096,