Update app.py

app.py

@@ -9,6 +9,7 @@ def split_model(model_path):
     from transformers import AutoConfig
     device_map = {}
     world_size = torch.cuda.device_count()
+    print(f"world_size:{world_size}")
     config = AutoConfig.from_pretrained(model_path, trust_remote_code=True)
     num_layers = config.llm_config.num_hidden_layers
     num_layers_per_gpu = math.ceil(num_layers / (world_size - 0.5))