diff --git a/lavis/models/blip_diffusion_models/modeling_ctx_clip.py b/lavis/models/blip_diffusion_models/modeling_ctx_clip.py
index 737b77d3f..e1d8bf6f8 100644
--- a/lavis/models/blip_diffusion_models/modeling_ctx_clip.py
+++ b/lavis/models/blip_diffusion_models/modeling_ctx_clip.py
@@ -13,9 +13,22 @@
 from transformers.models.clip.modeling_clip import (
     CLIPEncoder,
     CLIPPreTrainedModel,
-    _expand_mask,
 )
 
+def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None):
+    """
+    Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`.
+    """
+    bsz, src_len = mask.size()
+    tgt_len = tgt_len if tgt_len is not None else src_len
+
+    expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype)
+
+    inverted_mask = 1.0 - expanded_mask
+
+    return inverted_mask.masked_fill(inverted_mask.to(torch.bool), torch.finfo(dtype).min)
+
+
 
 class CtxCLIPTextModel(CLIPPreTrainedModel):
     config_class = CLIPTextConfig
diff --git a/requirements.txt b/requirements.txt
index 3f246993f..bf61fb500 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -20,11 +20,11 @@ scikit-image
 sentencepiece
 spacy
 streamlit
-timm==0.4.12
+timm
 torch>=1.10.0
 torchvision
 tqdm
-transformers==4.33.2
+transformers
 webdataset
 wheel
 torchaudio
@@ -35,5 +35,5 @@ peft
 
 easydict==1.9
 pyyaml_env_tag==0.1
-open3d==0.13.0
+open3d
 h5py