[shardformer] fix import (#5788)

colossalai/shardformer/modeling/llama.py

@@ -8,6 +8,10 @@ import torch.utils.checkpoint
 from torch import nn
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
 from transformers.cache_utils import Cache
+from transformers.modeling_attn_mask_utils import (
+    _prepare_4d_causal_attention_mask,
+    _prepare_4d_causal_attention_mask_for_sdpa,
+)
 from transformers.modeling_outputs import (
     BaseModelOutputWithPast,
     CausalLMOutputWithPast,
@@ -17,8 +21,6 @@ from transformers.models.llama.modeling_llama import (
     LlamaForCausalLM,
     LlamaForSequenceClassification,
     LlamaModel,
-    _prepare_4d_causal_attention_mask,
-    _prepare_4d_causal_attention_mask_for_sdpa,
     apply_rotary_pos_emb,
     repeat_kv,
 )

colossalai/shardformer/modeling/qwen2.py

@@ -9,13 +9,15 @@ from transformers.modeling_outputs import (
 )
 
 try:
+    from transformers.modeling_attn_mask_utils import (
+        _prepare_4d_causal_attention_mask,
+        _prepare_4d_causal_attention_mask_for_sdpa,
+    )
     from transformers.models.qwen2.modeling_qwen2 import (
         Qwen2Attention,
         Qwen2ForCausalLM,
         Qwen2ForSequenceClassification,
         Qwen2Model,
-        _prepare_4d_causal_attention_mask,
-        _prepare_4d_causal_attention_mask_for_sdpa,
         apply_rotary_pos_emb,
         repeat_kv,
     )