Open Fanqyu opened 10 months ago
Hi there, it's a great work. I'd like to know if there is any impact, such as slow training or inference speed, etc, for only using the "CPU version of multi-scale deformable attention".
Hi there, it's a great work. I'd like to know if there is any impact, such as slow training or inference speed, etc, for only using the "CPU version of multi-scale deformable attention".