why last layer self.conv_1x1_exp is Identity in code, but self.conv_1x1_exp is mapping to a higher dimension in paper?

[xuesongnie]

In code:
self.conv_1x1_exp = Identity()

In paper:

[shaktiwadekar9]

The table provides architecture details for MobileViTv3-XXS,XS, and S models [MobileViTv3-v1] and not for MobileViTv3-0.5,0.75 and 1.0 [MobileViTv3-v2]. The code part you are highlighting is present in the model definition for MobileViTv3-0.5,0.75 and 1.0. These models do not have conv-1x1 layer, therefore Identity() is used.

[xuesongnie]

The table provides architecture details for MobileViTv3-XXS,XS, and S models [MobileViTv3-v1] and not for MobileViTv3-0.5,0.75 and 1.0 [MobileViTv3-v2]. The code part you are highlighting is present in the model definition for MobileViTv3-0.5,0.75 and 1.0. These models do not have conv-1x1 layer, therefore Identity() is used.

Thanks！