GFT: Generative Fundamental Training

### Model description Hi, It is just an ad of our GFT. It has a novel attention head by fusing attention and mlp layer. It has an ultra-thin, ultra-deep architecture that maximizing model performance with minimal parameters. More interestingly, it equips a novel decoder call top-E(entropy) algorithm. Our model has 81 layers but only 300M parameters. Both Chinese and English(Biomedical) model are available. It is coded from scratch by cuda and C++, no DL framework is needed. Enjoy! ### Open source status - [X] The model implementation is available - [X] The model weights are available ### Provide useful links for the implementation https://github.com/wangyi-fudan/GFT Chinese Model Download 链接: https://pan.baidu.com/s/1HKw83YWttCIPdnZCZ-hdOA 密码: 0r1j English Medical Model Download 链接: https://pan.baidu.com/s/1yazmdB8xFzLwXbKmXWslRA 密码: rqdc