请问有没有 GPT2-ML 预训练模型，希望结合 DTR 进行微调

## 请求
实现MegEngine实现GPT2-ML(https://github.com/imcaspar/gpt2-ml)

## 任务描述
实验项目微调GPT2-ML，由于GPU资源不足（RTX 3060-12GB）无法调试，通过【机器之心】得知MegEngine1.4 DTR技术解决了动态图显存优化的问题，想请问有没有参考教程让我们能通过DTR实现对GPT2-ML的微调

## 目标
采用MegEngine实现的GPT2-ML模型，通过DTR进行微调

Start by reviewing the GPT2-ML project linked in the issue and the MegEngine DTR documentation or examples. Done means providing a MegEngine implementation of GPT2-ML together with a reference tutorial for fine-tuning it with DTR on the stated hardware constraints.

Written by the indexing model from the issue text.