Ke Meng comments

Repositories
Issues
Comments

Results 14 comments of


                                            Ke Meng

Implements spill and checkpoint functionalities in Vineyard

PR #815 fixes #740.

IPEX.LLM performance on Intel SPR with large batch size.

@jgong5 Thank you for your reply. I believe the CPU has the potential to replace the GPU as a cost-efficient solution for LLM inference. Although the GPU is fast, it...

[BUG] WCC failed on graphs with string oid

solved by https://github.com/alibaba/GraphScope/pull/2130

IPEX.LLM performance on Intel SPR with large batch size.

> For CPU (IPEX), this growth TREND appears abnormal when the batch size > 8, and it even decreases. Increasing the batch size actually becomes less efficient, which I think...