Continuing from trained data will start the output images at 00000 rather than their current value

[MaffooBristol]

Not sure if this is by design, but restarting from a checkpoint will start overwriting the images at index 00000.jpg or what-have-you. I feel that it should by default output at the next incremented value, such as 001234.jpg to avoid deleting existing images.

If the current implementation is by design, perhaps a config switch?

Cheers,
Matt

[martyn]

I've run into this to. It might not make it in 1.0 but I'll tag it so we can improve this after. Pytorch lets us save arbitrary variables to the save files so we can just add it there.