imporve log

Former-commit-id: a6abf375975ffea3d51e1b944c9855b5f62ffac8

src/llamafactory/model/model_utils/packing.py

@@ -41,9 +41,9 @@ from typing import TYPE_CHECKING, Tuple
 
 import torch
 import torch.nn.functional as F
-from transformers.utils.versions import require_version
 
 from ...extras import logging
+from ...extras.misc import check_version
 from ...extras.packages import is_transformers_version_greater_than
 
 
@@ -118,6 +118,6 @@ def configure_packing(model_args: "ModelArguments", is_trainable: bool) -> None:
     if not is_trainable or not model_args.block_diag_attn:
         return
 
-    require_version("transformers>=4.43.0,<=4.46.1", "To fix: pip install transformers>=4.43.0,<=4.46.1")
+    check_version("transformers>=4.43.0,<=4.46.1")
     transformers.modeling_flash_attention_utils._get_unpad_data = get_unpad_data
     logger.info_rank0("Using block diagonal attention for sequence packing without cross-attention.")