Open TheTinyTeddy opened 1 month ago
Hi, thank you for the great work!
I was wondering is DeepSpeed-Ulysses the sequence parallel method used in both inference/training of Open-Sora-Plan v1.2.0?
(As a side note, I think you could take a look at the hybrid ring/ulysses method featured in https://github.com/feifeibear/long-context-attention/tree/main )
We only used the sp method in the inference of Open-Sora-Plan v1.2.0. Thank you for your recommendation, we will pay attention to this work.
Hi, thank you for the great work!
I was wondering is DeepSpeed-Ulysses the sequence parallel method used in both inference/training of Open-Sora-Plan v1.2.0?
(As a side note, I think you could take a look at the hybrid ring/ulysses method featured in https://github.com/feifeibear/long-context-attention/tree/main )