You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
你的PR提交成功,感谢你对开源项目的贡献!
请关注后续CI自动化测试结果,详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.
Wanglongzhi2001
changed the title
[WIP] fmha support flash decoding
[Inference] fmha support flash decoding
Aug 8, 2024
Wanglongzhi2001
changed the title
[Inference] fmha support flash decoding
[Inference] Add multi-block mmha in FusedMT.
Aug 8, 2024
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PR Category
Inference
PR Types
New features
Description
pcard-71500
支持 FusedMT 中的 mmha 的 128K 上下文推理,开发了 multiblock 的 mmha kernel,期间使用 flash decoding 算法进行 head 的分块。
[NOTE]
单测里的上下文长度设置为 128K 可能导致单测耗时太长挂掉,因此单测的上下文长度仅设置成 2049 以及 layers 仅设置成 2 层。
[使用方式]
使用
export FLAGS_fused_multi_transformer_op_use_mbfmha =true
打开 FLAGS_fused_multi_transformer_op_use_mbfmha 后就行,无需额外操作,后续步骤和使用 FusedMT 的步骤一致。