MiniMind可视化扩展,展示模型结构、注意力机制和训练流程
Hello, first of all, thank you very much for your open-source MiniMind project. While studying the MiniMind code line by line, I made a visualization extension to make the model structure, attention mechanism, and training process more intuitive. Currently, it covers model architecture, MoE Block, GQA, RoPE, pretraining and fine-tuning, as well as DPO, PPO, GRPO, and other contents, and has been adapted for mobile devices. This project is based entirely on MiniMind itself, and does not include any components I added. I would like to ask you: If you think the content is appropriate, would you consider adding a link to the visualization project in the community learning resources section of the project README or in the recommended content collection in the Issues?
内容来源: jingyaogong/minimind