Add options and local inference

harubaru / waifu-diffusion

stable diffusion finetuned on weeb stuff

GNU Affero General Public License v3.0

1.94k stars 177 forks source link

Add options and local inference #42

Closed chavinlo closed 1 year ago

chavinlo commented 1 year ago

Added options to:

Disable Inference (it consumes about 2gb of VRAM even when not active)
Disable wandb

and:

if no hftoken is provided it just fills it with nothing so it doesn't argues
if wandb is not enabled, save the inference outputs to a local folder along with information about it

Note: I haven't tested this exact code but I did run a very (almost the same) similar one yesterday (with hivemind and other things) and it worked as expected

chavinlo commented 1 year ago

continuing the conversation we had on discord, would torch.cuda.empty_cache() be called inside the inference function?

chavinlo commented 1 year ago

done

harubaru commented 1 year ago

continuing the conversation we had on discord, would torch.cuda.empty_cache() be called inside the inference function?

Yeah but it's not too necessary for this PR it can be done later