Allocation on device error after one generation
Hello, thanks for this great checkpoint!
I'm having a little issue with it, I'm using the Mega v12.1 with the V3 workflow for I2V it runs quite well, but there is a annoying problem that everytime I generate one video, when I'm going to generate the second one it always fail with the following error
"Allocation on device This error means you ran out of memory on your GPU. TIPS: If the workflow worked before you might have accidentally set the batch_size to a large number"
After the error, if I run the workflow again it generates the video without any issue, but I have always to run the workflow twice.
I run it on vast.ai in a 4090.
Any tip to avoid this?
thanks in advance
edit: seems this happens only when I change my prompt, so I guess is some kind of cache?
I have not had that problem myself but I can offer some suggestions. Try using Shift + r or right click and clean up vram after you generate the first video or before running the second video. I have used a custom Node called Strawberry Fist vram optimizer also that will run automatically at what ever setting you set it to. There is nothing to connect to just have the node on your workflow set it and forget it. There may be other ways or nodes to do this but give those a try.
thank you Dredias11, I used the node you suggested and works like a charm!!
You are very welcome. So glad I could help!