Boning Cui

lol

oh you can host one on huggingface zerogpu like use gradio.server thingy expose restAPI or smth. then you can use your huggingface token to call it. its free

what model powers that

i should make my own bot too lol. just less mean

oh lol

for small models should the vocab also be small? because for my prototypes i used my 131K vocab one lol huge embedding

hmm the dude above is not wrong. 2K vocab is a bit small for a chat LLM. @Hoglet-33 Maybe for the next model consider increasing the vocab size a bit unless you know the tokenizer can very effectively split the text, or you could be wasting context :)

oh lol i though you were serious lol nvm

lol thats why it tried to roast me lol. how do you know what bot it is?

@ProCreations just out of curiosity, what optimizer do you usually use to train your ais? i use Muon