Dimension mismatch error when changing img_history_size in configs/base.yaml

[Wyatt2021]

When modifying the img_history_size parameter in the configs/base.yaml file, the model encounters a dimension mismatch error during the forward pass. Specifically, the error occurs when adding tensors img_c and self.img_cond_pos_embed.
Error Message:
The error encountered is as follows:

File "/path/to/RoboticsDiffusionTransformer/models/rdt/model.py", line 153, in forward
img_c = img_c + self.img_cond_pos_embed
RuntimeError: The size of tensor a (21870) must match the size of tensor b (4374) at non-singleton dimension 1
The shapes of the tensors involved:

img_c: torch.Size([2, 21870, 2048])
self.img_cond_pos_embed: torch.Size([1, 4374, 2048])

[soumajm]

Did you change the number of cameras in configs/base.yaml ?