react test evaluation config

[zhulinJulia24]

1. make pt_internlm2-chat-7b's max batch size to 64, becuase of OOM when max batch size is 128
2. react config template of internlm2、llama3 and qwen1.5
3. add config template of qwen2
4. add a choice by using local eval config
5. by using -r option when doing evaluation, if one model is failed we can rerun