Labmem009

请问大佬在CMeEE上跑，参数都没有改，为什么验证集只能跑到65%，测试结果只有61.5%。原先版本的GP也只到66.1%

1

用大佬git上的代码跑了CMeEE，参数都没有改，但是验证集只能跑到65%，测试结果只有61.5%。原先版本的GP也只到66.1%，这是为啥啊QAQ我用的也是roberta-zh-large-bert

请问是否支持200k上文的微调，需要什么样的配置？

5

目前我使用的是zero3对interlm2-chat-20b进行全量微调，8*A100只能微调2k上文，请问200k或者比较长几十k的上文微调需要怎么设置？

作者您好，我在blog中看到你对四个方法在CMeEE上的表现都做了对比，但是代码实现里用的不是阿里云标注比赛提供的原数据格式，请问Global Pointer 和其它NER方法所使用数据是否有处理脚本或者数据范例以供参考？

I want to apply SPIN method on llama2 with alpaca-like finetuning datasets. What changes should I make to apply the SPIN method? Thanks a lot!

Do you have any plan to release new model weight based on LLaVA1.5？ For LLaVA1.5-7B is still weak in OCR, I think maybe your repo can help

### Question My environment 'llava' has been running well with LLaVA and LLaVA-Plus, but it broke down with CUDA error after I run pip install clip-retrieval. What are the correct...

SPAE code release

9

Great work! After reading your paper _SPAE: Semantic Pyramid AutoEncoder for Multimodal Generation with Frozen LLMs_ , I'm very interested in the implementation, especially how the image is reconstructed from...

I want to apply Self-rewarding and SPIN method on llama2 with alpaca-like finetuning datasets. What changes should I make to apply the method? And what config should I use? Thanks...

In jp.jsonl, all the question is "null", is there any mistake?

3

Some lines are "T2T" type, but still have no question, is there any mistake?