We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
1 parent dc74199 commit e4e13f5Copy full SHA for e4e13f5
official/cv/segment-anything/segment_anything/modeling/prompt_encoder.py
@@ -56,7 +56,10 @@ def __init__(
56
self.no_mask_embed = nn.Embedding(1, embed_dim)
57
58
self.text_embeddings = nn.Embedding(1, embed_dim)
59
- self.text_proj = nn.Dense(in_channels=text_feature_dim, out_channels=embed_dim)
+ # self.text_proj = nn.Dense(in_channels=text_feature_dim, out_channels=embed_dim)
60
+
61
+ self.text_proj = nn.Dense(in_channels=text_feature_dim, out_channels=embed_dim) \
62
+ if text_feature_dim != embed_dim else nn.Identity() # clip need projection while blip2 not
63
64
def get_dense_pe(self) -> ms.Tensor:
65
"""
0 commit comments