jyshee

Results 4 comments of jyshee

high similarity between abnormal face images

Do you mean the threshold of a certain FPR? Then the threshold is high because of such abnormal fp imgs.

High training loss of LLaMA 13B

It seems that you should add `"gradient_accumulation_steps": "auto"` to deepspeed config, otherwise, the gradient_accumulation_steps in deepspeed is still 1

官网上每个数据集所使用的config文件的版本如何知道？

> For datasets on OpenCompass 1.0 Leaderboard, you can just move your cursor on the score of the dataset to find the config. For example, the config for C-Eval can...

DeepSpeed sequence parallelism (aka Ulysses) integration with HF transformer

@samadejacobs so how is this going now?