New model, please add: Nexus-Erebus-135M

#54
by MaliosDark - opened

Nexus-Erebus-135M, 135M from scratch. 0-shot, lm-eval (acc_norm for hellaswag/arc, acc for piqa):

hellaswag 30.61
arc-easy 48.61
arc-challenge 25.43
piqa 64.53
arithmark2 66.68
avg 49.71

repo: MaliosDark/Nexus-Erebus-135M

same digit-level tokenizer as always, that's why the math holds up. thanks!

Hi @MaliosDark , awesome new model, currently don't have the equipment to update the actual doc, could you open this as a pr and add the model and I would be happy to merge it

isnt this a specialist?

Axiomic Labs org

Algorithm will decide

opened the PR with the model added: #55. thanks!

Axiomic Labs org

Merged, TY!

Datdanboi25 changed discussion status to closed

Sign up or log in to comment