Closed YanShuang17 closed 1 year ago
Hi @YanShuang17 , this difference is caused by the different implementations of deformable attention in the original MaskDINO repo and detrex. In detrex, the deformable attention code contains adding positional queries and residual connection.
OK. got it, thx
Location:
xxx/modeling/pixel_decoder/maskdino_encoder.py
, in forward function ofMSDeformAttnTransformerEncoderLayer
:MaskDINO
implementation:detrex
implementation(residual is lost):@HaoZhang534