Muhammad Rizqi Nur

You can use the prompt editing/scheduler syntax [from:to:when]. For something major that needs to take shape, use from with low when. For example, [dress::0.2]. Something that's already in the picture...

> I also found this... using cloudflare tunnel, its a bit sketchy > > ``` > !wget -q https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64.deb && dpkg -i cloudflared-linux-amd64.deb > !python webui.py --autolaunch --precision autocast &...

Nvm it's actually caused by StylePile for some reason

> I am getting a very similar warning as described in #978 when running with multiiprocessing: `Loky-backed parallel loops cannot be called in a multiprocessing, setting n_jobs=1`. I use scikit-learn's...

You can use 0.6.0, but the hardcoded dataset link in 0.3.1 - 0.6.0 is dead. I patched it up a bit here https://github.com/R-N/torchtext/tree/0.6.1

Yeah, the None option should revert it to the VAE bundled with the checkpoint, so this is a bug, indeed. This PR was just cherry picked and I havent had...

Yep, I confirm that the None option is indeed bugged.

I got it. I don't know what happened, but it's as mentioned in #4283. Model caching now requires deepcopy somehow. I know for sure that this wasn't the case, because...

Yeah, please try it and confirm if it works now

Hmm it worked for me (included test in the first post). I wonder what was wrong? It seems that your branch is similar with deepcopy, and instead of using loaded_vae_file...