lets put an actual model on here
#4
by sethtadd - opened
lol our computers keep exploding
This PR works great for me, using oobabooga/text-generation-webui.
Barely run on 3080Ti. Nice work!
in order to be compatible with the hf model downloader download-model.py that comes with oobabooga/text-generation-webui, the contents of the PR would need to go into the root
folder, and not the 7B-2nd-train.
Ideally, you would just overwrite the files each epoch you train, and the git will let anyone go grab the older versions. no reason to have them in separate folders.
8bit-coder changed pull request status to merged