Fused_mt Branch Migration #64125
Merged
Fused_mt Branch Migration #64125
Conversation
This file contains 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! |
|
|
|
|
penPenf28
requested review from
wanghuancoder,
luotao1,
Aurelius84,
XiaoguangHu01 and
qili93
as code owners
paddle/fluid/operators/fused/fused_multi_transformer_helper.cu.h
Outdated
Show resolved
Hide resolved
sneaxiy
approved these changes
May 22, 2024
XieYunshen
approved these changes
May 23, 2024
yuanlehome
approved these changes
May 23, 2024
qingqing01
approved these changes
May 23, 2024
YuanRisheng
approved these changes
May 23, 2024
wanghuancoder
approved these changes
May 23, 2024
chen2016013
pushed a commit
to chen2016013/Paddle
that referenced
this pull request
May 26, 2024
* Merge fused_mt branch * Adjusted fuse_mt_int8 * Revert attention_layer_norm.h * Revert paddle/phi/kernels/fusion/gpu/fmha_ref.h * Add win support and refine format. * Reformat for win. * Removed redundant files, now only supports flash_attn_v2 and variable length * Refine static_fused_ft test * Refine fused_mt related testcase * Remove custom_adll_reduce * Remove operator cublaslt and revert parallel test * Refine empty seq_len * Refine ft * Refine ft_static test * Remove float32 support and static parallel ft test * Refine type static error. * Fix doc type error * Fuse_mt code format * Remove some redundant code * Remove redundant attention_layer_norm.h * Remove redundant code in ft_op * Remove Redundant code and skip fuse_mt doctest * Remove redundant fmha_ref mmha_util and other code * Remove redundant kernel * Remove redundant file * Refine fuse_mt code * Refine cublaslt comment
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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
移植如下内容，并做了一定程度上的适配
重点修改paddle/fluid/operators/fused/fused_multi_transformer_op.cu，增加了GQA支持，目前仅支持flash_attention_v2的底层实现（only float16/bfloat16）
本地单测，移除了padding输入为定长的测试，只支持variable长度