diff --git a/internlm/model/modeling_llama.py b/internlm/model/modeling_llama.py
index f489f36..a5362c4 100644
--- a/internlm/model/modeling_llama.py
+++ b/internlm/model/modeling_llama.py
@@ -44,7 +44,7 @@ try:
     from flash_attn.modules.mlp import ParallelFusedMLP
     from flash_attn.ops.layer_norm import dropout_add_layer_norm
 except ImportError:
-    raise ImportError("Please check your flash_attn version >= 2.0.0.")
+    pass
 
 MODEL_TYPE = "LLAMA"