[Paddle Inference] Fix mmha when src_mask is not equal to zero #57936
Conversation
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! |
carryyu
reviewed
Oct 9, 2023
| for (int ii = 0; ii < K_VECS_PER_THREAD; ++ii) { | ||
| int jj = ii * params.max_seq_length + ti; | ||
| if (ti < act_time_step) { | ||
| if (ti < act_time_step + 1) { |
|
|
||
| // bool is_mask = false; | ||
| if (ti < act_time_step && tid % THREADS_PER_KEY == 0) { | ||
| if (ti < act_time_step + 1 && tid % THREADS_PER_KEY == 0) { |
There was a problem hiding this comment.
这里的加1相关的逻辑 是否可以放到613行，加入一个判断，因为大多数情况下当前位置都是不会加mask的。
xiaoxiaohehe001
changed the title
carryyu
previously approved these changes
Oct 10, 2023
carryyu
reviewed
Oct 10, 2023
| // bi * (params.timestep + 1) + params.timestep]; | ||
| // qk += static_cast<float>(mask); | ||
| qk *= params.inv_sqrt_dh; | ||
| auto mask_bhi = params.mask_broadcast_num_heads ? bi : bhi; |
There was a problem hiding this comment.
这部分可以放到if分支里面，当有attn mask的时候才计算mask_bhi
carryyu
approved these changes
Oct 11, 2023
Frida-a
pushed a commit
to Frida-a/Paddle
that referenced
this pull request
Oct 14, 2023
…ePaddle#57936) * fix_mmha_scrmask * fix_mmha_scrmask * remove_mask_to_qk_smem_act_time_step * remove_add_! * mask_bhi
jiahy0825
pushed a commit
to jiahy0825/Paddle
that referenced
this pull request
Oct 16, 2023
…ePaddle#57936) * fix_mmha_scrmask * fix_mmha_scrmask * remove_mask_to_qk_smem_act_time_step * remove_add_! * mask_bhi
danleifeng
pushed a commit
to danleifeng/Paddle
that referenced
this pull request
Nov 14, 2023
…ePaddle#57936) * fix_mmha_scrmask * fix_mmha_scrmask * remove_mask_to_qk_smem_act_time_step * remove_add_! * mask_bhi
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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 types
Others
PR changes
Others
Description
Fix mmha diff when src_mask is not equal to zero.
Pcard-71502