Update app.py

app.py

@@ -60,10 +60,10 @@ accelerator = Accelerator()
 
 model = AutoModelForCausalLM.from_pretrained(model_id, token= token, 
                                                  # torch_dtype= torch.uint8, 
-                                             torch_dtype=torch.float32,
+                                             torch_dtype=torch.bfloat16,
                                               # load_in_8bit=True,
                                              # #  # torch_dtype=torch.fl,
-                                             # attn_implementation="flash_attention_2",
+                                             attn_implementation="flash_attention_2",
                                              # low_cpu_mem_usage=True,
                                              # device_map='cuda',
                                              # device_map=accelerator.device_map,