55 ms·
In my experience, Wizard-Vicuna output at 13b will match or exceed the prose quality of GPT3.5 in smaller generation tasks. I've been using it for a Discord bot
by smoldesu 3y ago
In my experience, Wizard-Vicuna output at 13b will match or exceed the prose quality of GPT3.5 in smaller generation tasks. I've been using it for a Discord bot to great success.