Ye Ting

### Feature request 我用[Fastchat](https://github.com/lm-sys/FastChat)里面的server加载aquila-7b模型时，提示报错，能否支持一下采用Fastchat模型加载方式？ 报错信息如下： ```bash (base) root# curl http://localhost:30010/v1/completions -H "Content-Type: application/json" -d '{ "model": "aquila-7b", "prompt": "北京在哪？", "max_tokens": 1000, "temperature": 0.5}' {"object":"error","message":"**NETWORK ERROR DUE TO HIGH TRAFFIC. PLEASE...

add argument `--rotary-base` for gpt model

1

add argument `--rotary-base` for gpt model

[QUESTION] What is the difference between with/without mcore model in pretrain_gpt.py?

2

What is the difference between with/without mcore model in pretrain_gpt.py? [pretrain_gpt.py#L33](https://github.com/NVIDIA/Megatron-LM/blob/5f9c870f9f24b482509699d206a9dbb00958f6fc/pretrain_gpt.py#L33)

add megatron_mcore saver/loader

12

add saver/loader for mcore checkpoint convert