diff --git a/trainer/train_full_sft.py b/trainer/train_full_sft.py
index 9b04d5e..2409ba3 100644
--- a/trainer/train_full_sft.py
+++ b/trainer/train_full_sft.py
@@ -97,7 +97,7 @@ def init_model(lm_config):
     tokenizer = AutoTokenizer.from_pretrained('../model')
     model = MiniMindForCausalLM(lm_config)
     moe_path = '_moe' if lm_config.use_moe else ''
-    ckp = f'{args.save_dir}/full_sft_{lm_config.hidden_size}{moe_path}.pth'
+    ckp = f'./out/pretrain_{lm_config.hidden_size}{moe_path}.pth'
     state_dict = torch.load(ckp, map_location=args.device)
     model.load_state_dict(state_dict, strict=False)