Improved efficiency in the forward part of the decoder.

[tyui592]

ConditionalDETR/models/transformer.py

Line 145 in 8ea7c3f

obj_center = reference_points[..., :2].transpose(0, 1) # [num_queries, batch_size, 2]

Since reference_points does not change in the for loop, it is more efficient that pre-calculate the obj_center.