Lnyan

@undeathable I made a dirty hack for GPUs with 12/24G vRAM: https://github.com/lkwq007/CogView2-low-vram

> > @undeathable I made a dirty hack for GPUs with 12/24G vRAM: https://github.com/lkwq007/CogView2-low-vram > > Thanks! https://github.com/lkwq007/CogView2-low-vram works with --single-gpu perfectly solved my original issue! The only weird thing...

> @lkwq007 Could you pleae briefly explain how to use without `fused_layer_norm` from apex . I have installed apex yet. Thank you! @Limbicnation I suspect that `"RuntimeError: Expected all tensors...

 It also requires prompt engineering. (here I use "Ivan Grohar, impressionist, painting style") For high resolution image, you may need to resize the image or increase the size of...

> > It also requires prompt engineering. (here I use "Ivan Grohar, impressionist, painting style") For high resolution image, you may need to resize the image or increase the size...

Thanks for your suggestions. Lags can be avoided by using a local server. I'll look into how to optimize the performance in colab. Undoing one step might be easy to...

commit/retry/undo added uploading new images will now clear the canvas

@arpitest I have ported the canvas interaction to frontend via pyscript and gradio. The colab version should work without lags now. @fractaltr33 Erasing is also implemented with the updated version.

> Suggestion: It would be nice to be able to change the model (model.ckpt) that Stable Diffusion uses, if possible The current implementation relies on `diffusers`. You may change the...

> `from_pretrained` is a method of `StableDiffusionPipeline` ``` text2img = StableDiffusionPipeline.from_pretrained( "path_to_model_folder", revision="fp16", torch_dtype=torch.float16 ).to("cuda") ```