Hello,
I am attempting to fine-tune a MaskDINO model using a Swin-B backbone and pre-trained weights. I was using the configuration files provided in the repository to set up my training script.
I noticed an inconsistency in the directory located at:
configs/coco/instance-segmentation/swin/
The files within this directory are named maskdino_R50_...yaml, which define a ResNet-50 architecture, rather than the expected Swin Transformer architecture.
Could you please upload the correct configuration files that define the Swin-B/Swin-L backbones?
Providing these files would greatly help users who are trying to match the Swin directory name with the appropriate configuration content, preventing architecture mismatch errors during model setup.
Thank you for your assistance.
Hello,
I am attempting to fine-tune a MaskDINO model using a Swin-B backbone and pre-trained weights. I was using the configuration files provided in the repository to set up my training script.
I noticed an inconsistency in the directory located at:
configs/coco/instance-segmentation/swin/
The files within this directory are named maskdino_R50_...yaml, which define a ResNet-50 architecture, rather than the expected Swin Transformer architecture.
Could you please upload the correct configuration files that define the Swin-B/Swin-L backbones?
Providing these files would greatly help users who are trying to match the Swin directory name with the appropriate configuration content, preventing architecture mismatch errors during model setup.
Thank you for your assistance.