Insufficient video memory

[userLx888]

Training structure_generator.In order to solve the problem of insufficient video memory during generator, reducing batch_size leads to an error when saving the training results : Index 1 is out of bounds for axis 0 with size 1.I hope I can get your help. Thank you!

[USTC-JialunPeng]

What is your batch size? And please attach your error log.

[userLx888]

my batch size is 2,

[USTC-JialunPeng]

I found the problem. The visualization of structure feature maps shows 4 images. You can replace line162 with height = size and line167 with for i in range(1): to show only 1 image.

Diverse-Structure-Inpainting/net/nn.py

Lines 155 to 179 in 0def26d

	def structure_visual(gt, masked, recons_gen, recons_gt, iter, size, save_dir):
	""" Show 4 generated structure feature maps in the training of structure generator."""
	# gap between each images
	gap = 2
	# height and width of result image
	height = size * 4 + gap * 3
	width = size * 4 + gap * 3
	result = 255 * np.ones((height, width, 3), dtype=np.uint8)
	for i in range(4):
	gt_i = ((gt[i] + 1.) * 127.5).astype(np.uint8)
	masked_i = ((masked[i] + 1.) * 127.5).astype(np.uint8)
	recons_gen_i = ((recons_gen[i] + 1.) * 127.5).astype(np.uint8)
	recons_gt_i = ((recons_gt[i] + 1.) * 127.5).astype(np.uint8)
	# fill the images into grid
	result[i*(size+gap):i*(size+gap)+size, 0*(size+gap):0*(size+gap)+size, ::-1] = gt_i
	result[i*(size+gap):i*(size+gap)+size, 1*(size+gap):1*(size+gap)+size, ::-1] = masked_i
	result[i*(size+gap):i*(size+gap)+size, 2*(size+gap):2*(size+gap)+size, ::-1] = recons_gen_i
	result[i*(size+gap):i*(size+gap)+size, 3*(size+gap):3*(size+gap)+size, ::-1] = recons_gt_i
	cv2.imwrite(os.path.join(save_dir, 'structure%d.png' % iter), result)

[userLx888]

OK OK,I get it ,thank you very much!

[userLx888]

OK OK,I get it ,thank you very much!