Training for a single machine with multiple GPUs

[chenjiahao0124]

Hello, you have done a great job. May I ask how to set up training for a single machine with multiple GPUs? I encountered the following error while conducting multi GPU training and hope you can answer it. Thank you.

File "\models\graspnet.py", line 107, in forward
grasp_top_views_rot, _, _, , end_points = match_grasp_view_and_label(end_points)
File "\utils\label_generation.py", line 142, in match_grasp_view_and_label
top_template_views_rot = torch.gather(template_views_rot, 2, top_view_inds).squeeze(2)
RuntimeError: Size does not match at dimension 3 expected index [1, 1024, 1, 3, 3] to be smaller than src [1, 1024, 300, 1, 3] apart from dimension 2