360 has open sourced 360 Smart Brain 7B (7 billion parameter model) on GitHub. The model uses 3.4 trillionTokenscorpus for training, including Chinese, English and code, opening up the model for three different text lengths.360 said it validated the model performance on multiple review datasets and achieved good results. In addition to model weights, tools such as fine-tuning training code and inference code are also open-sourced.
Open source address:
https://github.com/Qihoo360/360zhinao