search

intel / xFasterTransformer

Apache License 2.0
270 stars 53 forks source link

[Layers] Add qwenRope support for Qwen1.0 in CB mode #449

Closed abenmao closed 3 weeks ago

changqi1 commented 3 weeks ago

I think this kernel APIs on continuous batching version and on continuous batching version are the same. Next step. we could merge two into one kernel API. OK for current version.

abenmao commented 3 weeks ago

I think this kernel APIs on continuous batching version and on continuous batching version are the same. Next step. we could merge two into one kernel API. OK for current version.

Yes, maybe we can remove the older version in the next step.