BeanLabs/ BeansLLM-2-M:latest

3 2 weeks ago

BeansLLM 2 mid training release with zero finetuning, loss 2.9

... /
params
421d90abf82e ยท 103B
{
"num_ctx": 2048,
"repeat_penalty": 1.3,
"stop": [
"<|endoftext|>"
],
"temperature": 0.7,
"top_p": 0.9
}