Add UC Berkleys Large World Models
Can someone add support for the Large World Models(Text and Multimodal) from UC Berkley? https://largeworldmodel.github.io/
The text models seem to just be llama so you can use convert.py for those
The interesting part is probably the implementation of ring attention
The interesting part is probably the implementation of ring attention
I haven't totally understood it, but would it help with memory consumption on large contexts?
As far as I understood it ring attention is an efficient way to compute attention in many GPUs
This issue was closed because it has been inactive for 14 days since being marked as stale.