Update app.py

app.py

@@ -4,19 +4,23 @@ from llama_cpp import Llama
 from huggingface_hub import hf_hub_download
 import random
 
-# Initialize model
-model_path = hf_hub_download(
-    repo_id="AstroMLab/AstroSage-8B-GGUF",
-    filename="AstroSage-8B-Q8_0.gguf"
-)
+@spaces.GPU
+def initialize_model():
+    model_path = hf_hub_download(
+        repo_id="AstroMLab/AstroSage-8B-GGUF",
+        filename="AstroSage-8B-Q8_0.gguf"
+    )
+    
+    llm = Llama(
+        model_path=model_path,
+        n_ctx=2048,
+        chat_format="llama-3",
+        n_gpu_layers=-1,  # ensure all layers are on GPU
+        flash_attn=True,
+    )
+    return llm
 
-llm = Llama(
-    model_path=model_path,
-    n_ctx=2048,
-    chat_format="llama-3",
-    n_gpu_layers=-1,  # ensure all layers are on GPU
-    flash_attn=True,
-)
+llm = initialize_model()
 
 # Placeholder responses for when context is empty
 GREETING_MESSAGES = [