mradermacher

/

model_requests

like 49

They are getting bigger... Queued :)

haha thanks

bro, stop quantizing that, the model is not good, i've detected too much problems in the 101b model, the other is going great but the 101b upscaled is not good

you can delete it

I tried them booth and can confirm that Guilherme34/Reasoner-Llama3.1-70b is awesome as it has chain of thought finetuned into it while Guilherme34/Reasoner-101B is unfortunately the exact opposite and after a few tokens starts to endlessly repeat itself in a loop and sometimes even starts with "thinks2" instead of "thinks" breaking the entire chain of thought chain.

Good good, it's gone :)