Chujie Zheng comments

Results 9 comments of


                                            Chujie Zheng

Z97M-D3H有升级到Big Sur或者Monterey的计划吗？

多谢建议~ 我已经参照网上的教程把我的台式机顺利升到了Monterey，主要步骤是： - 定制usb - 更新驱动 - 更新opencore和config.plist文件

2021-7-4 catalina 升级到 10.15.7 小升级

HDMI+DVI双屏感觉不太稳定，准备淘个免驱卡接双DP了

Why doesn't the model input include attention_mask?

Because GPT is a uni-directional language model. It does not need attention mask.

Got stuck when I trying to load a dataset

Same here. Is there any solution?

(QLoRA) DPO without previous SFT

I guess mistral before SFT is just a base language model, the chat template is used for the instruction-tuned version (mistral-instruct). On the other hand, the LLM leaderboard does not...

运行bash RUN/infer_vanilla.sh报错

检查下路径，看起来像是你的bash脚本里no_repeat_ngram_size参数后没有留空格

BlenderEmotionalSupport.py代码有误

@lsy641

tr_loss1, logging_loss1 = 0.0, 0.0

@lsy641

Slow inference performance for large Llama models compared to naive MP

Also found that 4*gpus tp is much slower than 2*gpus tp, while the latter is still a bit faster than 2*gpus pp.