Peng

I am always a student for AI System

Results 15 comments of


                                            Peng

[Bug] 使用上万亿高质量【预料】，建立模型超强知识体系；

fixed in https://github.com/InternLM/InternLM/pull/73

Some documentation is missing

We are doing this one in https://github.com/InternLM/InternLM/pull/245

[Bug] Python3.10环境运行加载模型错误

@exceedzhang .to(torch.bfloat16)后是否还会OOM?

在安装docker环境时，总是爆出这个错误，无法解决

@JiaoPL

升级CUDA版本以支持Windows版本的flash-attention

我们来计划一下

[Feature] support sequence parallel in head layer and embedding layer

or at least they should not impact the value of global norm

[Bug] do not use torch.cuda.current_device() as device, since it only retures an int

![Uploading 企业微信截图_f6b71604-f3a9-4917-b0f4-e36960e097c3.png…]()

[Bug] do not use torch.cuda.current_device() as device, since it only retures an int

[QA] Internevo是否支持tied_embedding?

@zigzagcai 帮忙看一下

[Feature] 不使用memory pool

@blankde @huangting4201

1
2
›