Decrease memory needs
carlosedubarreto opened this issue · comments
I have a simple RTX 2060, with 6gb, is that possible to run on it?
I'm testing with the tanksandtample data, but its complaining of gpu memory.
but if I decrease the resolution in the arguments bwhen running its seems to doesnt make difference. If I change the ndepth it does work, to a certain point, as it complains
RuntimeError: The size of tensor a (3) must match the size of tensor b (4) at non-singleton dimension 2
I think the problem is because the training checkpoint provided was trained using the default depths 48,32,8
So in the and I cant run the tests.
Is there another way?
Thanks a lot.
I do not think this is a memory issue. I ran into a similar error when training a CNN in the past. I was trying to feed a spectrogram (an image) of the wrong size to the model. Since I was working with audio, I just had to change the sample rate/bit depth (resolution).
if you read the error backwards; the 2 dimensional input must have axis of the same size.
The datasets that they used were preprocessed, so you might need to do the same if you want to use custom images for training.
If ram is the issue, pruning might help. here is an example of pruning that was used to slim-down the stable-diffusion model & remove redundant data for their use case (anime):
https://github.com/harubaru/waifu-diffusion/blob/main/scripts/prune.py
Thans a lot for the answer. I just found out a way to run on my 6gb card.
decrease the max_h/max_w size and the numdepth
This one works for me
python eval_rcmvsnet_dtu.py --numdepth 128 --max_h 512 --max_w 512
No problem! I'm glad that I could help :)
PS: @carlosedubarreto, please consider adding linux support for your blender plugin <3
I have a few (in reality there are at least five people asking) people asking for it too, but linux support is a hard one, because I hardly have time to make the windows ones 😅, and I'm not used to linux system. windows is my home 😃
@carlosedubarreto I think I saw a discord link somewhere, I'll hop in because I'd love to help & not spam this thread.