So if I understand this right, textual_inversion creates prompts for a series of images such that you can use them as inputs for Stable Diffusion. We have a massive dataset of marseys, so maybe if you fed them in as an input, you could get a prompt that describes marsey as well as possible, allowing for unlimited marsey generation, rendering chiobu obsolete once and for all!
I would try myself but not only am I a Windows-cel, I'm also a GPU-let, using AMD in an NVIDIA user's world. Probably going to try either dual booting into ubuntu or something like that later, but that will take a while of finangling, so I thought I'd suggest it to one of the many nerds who probably already have done this
Jump in the discussion.
No email address required.
yeah I was looking into it but I don't have a GPU cause I'm not a gaymer lmao
you could try this too: Run Stable Diffusion on Intel CPUs
really pretty cool
https://textual-inversion.github.io/
Edit: oh ya @HeyMoon I ran some interrogator thingy on @brzl's marsey commission and got this
Jump in the discussion.
No email address required.
hmm that is interesting, I have never heard of an interrogator before, but it makes sense
Jump in the discussion.
No email address required.
More options
Context
i wonder how many years until TPUs are outdated or available for consumer machines
Jump in the discussion.
No email address required.
The only TPU I used were Google Colab ones and they honestly sucked butt everytime, somehow almost worse than cpu lmao
Are they good in general or kinda meh compared to current GPUs?
Jump in the discussion.
No email address required.
idk im not a machinelearningcel but the really expensive ones are probably faster than equivalent non focused stuff for those tasks?
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I think some SBCs have TPUs already: https://brainjar.ai/blogs/coral-tpu-jetson-nano-performance
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context