vlad.andronik
vlad.andronik
It is possible only to use light version in Colab. I tried to increase the number of parameters. But for my task the results were unpleasant, it seems that only...
@lllyasviel Have you tried training the ControlNet on some version of optical flow for video2video applications? There was a discussion where you asked which models are most wanted for the...
@etiennedub it might be the regions of occlusions, i.e. when the head is shifted, we need to inpaint what is behind etc. It is usually computed by taking two optical...
Thank you, so I would also need 70GB of video memory? Would it be possible to load it into 24GB with quantization?
Did you investigate if it affects performance too much, additional hallucinations?
Thank you for the response!
@chanel111 is it possible to use VL-plus as non-Chinese?
@chanel111 I meant if I could use it through API, if I am non-Chinese, because I had problems with alibaba cloud setup.
> @asomoza Most XL models and in particular the Lightning models behave rather strangely with the img2img pipelines, which needs a separate investigation but the timestep issue is still very...