This repository hosts a [Janus-Pro 1B] trained by GCPO. The reward model is Geneval.

The training code is available at GCPO

Safetensors

Model size

2B params

Tensor type

BF16

Model tree for zghhui/Janus-Pro-1B-GCPO-Geneval

Base model

Finetuned

(5)

this model

Collection including zghhui/Janus-Pro-1B-GCPO-Geneval