parallel transformer

[raghukiran1224]

Hi,

Any advice on how I can go about enabling a PaLM style transformer architecture using DeepSpeed, specifically for inference only?

Thanks!