diff --git a/bmtrain/nn/burst_utils.py b/bmtrain/nn/burst_utils.py
index 413100f..ebd9099 100644
--- a/bmtrain/nn/burst_utils.py
+++ b/bmtrain/nn/burst_utils.py
@@ -1,11 +1,5 @@
 import bmtrain as bmt
 import torch
-from flash_attn.flash_attn_interface import (
-    _flash_attn_forward as _flash_attn_forward_cuda,
-)
-from flash_attn.flash_attn_interface import (
-    _flash_attn_backward as _flash_attn_backward_cuda,
-)
 import inspect
 
 class ops_wrapper:
@@ -228,6 +222,9 @@ def inter_flash_attn_backward_triton(
 
 
 def inter_flash_cuda_fwd(q, k, v, o, lse, softmax_scale=1.0, causal=False):
+    from flash_attn.flash_attn_interface import (
+        _flash_attn_forward as _flash_attn_forward_cuda,
+    )
     o_i, _, _, _, _, lse_i, _, _ = _flash_attn_forward_cuda(
         q,
         k,
@@ -250,6 +247,9 @@ def inter_flash_cuda_fwd(q, k, v, o, lse, softmax_scale=1.0, causal=False):
 def inter_flash_cuda_bwd(do, q, k, v, o, lse, dq, dk, dv, softmax_scale, causal):
     dk_ = torch.empty_like(q)
     dv_ = torch.empty_like(q)
+    from flash_attn.flash_attn_interface import (
+        _flash_attn_backward as _flash_attn_backward_cuda,
+    )
     if len(o.shape) == 3:
         # use sum(o_i * gradoutput) as delta and pass a empty out to flash backward
         # this feature requires a build of this PR: https://github.com/Dao-AILab/flash-attention/pull/905