Update merge_lora_params.py

Author: Galaxy1458

llm/merge_lora_params.py

@@ -110,7 +110,7 @@ def merge():
             model = AutoModelForCausalLM.from_pretrained(
                 lora_config.base_model_name_or_path,
                 config=config,
-                low_cpu_mem_usage=args.low_gpu_mem,
+                low_cpu_mem_usage=True,
             )
             model = LoRAModel.from_pretrained(model=model, lora_path=args.lora_path, lora_config=lora_config)
         model.eval()
@@ -125,7 +125,7 @@ def merge():
         model = AutoModelForCausalLM.from_pretrained(
             lora_config.base_model_name_or_path,
             config=config,
-            low_cpu_mem_usage=False,
+            low_cpu_mem_usage=args.low_gpu_mem,
         )
         lora_config.merge_weights = True
         model = LoRAModel.from_pretrained(model=model, lora_path=args.lora_path, lora_config=lora_config)