Merge pull request #476 from dlutsniper/main

Update finetune.py For lower cpu memory

finetune.py

@@ -278,7 +278,7 @@ def train():
 
     local_rank = training_args.local_rank
 
-    device_map = None
+    device_map = "auto"
     world_size = int(os.environ.get("WORLD_SIZE", 1))
     ddp = world_size != 1
     if lora_args.q_lora:
@@ -302,6 +302,7 @@ def train():
         config=config,
         cache_dir=training_args.cache_dir,
         device_map=device_map,
+        low_cpu_mem_usage=True if training_args.use_lora and not lora_args.q_lora,
         trust_remote_code=True,
         quantization_config=GPTQConfig(
             bits=4, disable_exllama=True