Update app.py

app.py

@@ -2,14 +2,16 @@ from peft import PeftModel
 from transformers import LlamaTokenizer, LlamaForCausalLM, GenerationConfig
 import gradio as gr
 from torch.cuda import is_available
+from torch import bfloat16
 
-if True:#is_available():
+if is_available():
     options = dict(
         load_in_8bit=True,
         device_map="auto",
     )
 else:
     options = {
+        "dtype": bfloat16
         #"low_cpu_mem_usage": True,
         #"offload_state_dict": True,
         #"offload_folder": "offload",