Upload app.py with huggingface_hub

app.py

@@ -10,7 +10,7 @@ tokenizer = AutoTokenizer.from_pretrained(model_name)
 print("Loading model (this may take a while on CPU)...")
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    torch_dtype=torch.float32,
+    torch_dtype=torch.float16,
     low_cpu_mem_usage=True,
 )
 model.eval()