Update app.py

app.py

@@ -60,7 +60,7 @@ if not torch.cuda.is_available():
 
 if torch.cuda.is_available():
     model_id = "PartAI/Dorna-Llama3-8B-Instruct"
-    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
+    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", torch_dtype=torch.bfloat16)
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     
 generation_speed = 0