Update app.py

app.py

@@ -10,7 +10,7 @@ peft_model_id = "Pr123/TinyLlama-EA-Chat"
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model = AutoPeftModelForCausalLM.from_pretrained(
     peft_model_id,
-    torch_dtype=torch.float16  # Keeping half precision
+    torch_dtype=torch.bfloat16
 ).to(device)
 
 tokenizer = AutoTokenizer.from_pretrained(peft_model_id)