【Inference PIR】add fused_rotary_position_embedding_pass #65265
Merged
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! |
yuanlehome
reviewed
Jun 18, 2024
| @@ -623,6 +623,7 @@ const std::vector<std::string> kPirGpuPasses{ | |||
| "transpose_flatten_concat_fuse_pass", | |||
| "remove_redundant_transpose_pass", | |||
| "transfer_layout_pass", | |||
| "fused_rotary_position_embedding_pass", | |||
There was a problem hiding this comment.
这个pass放在embedding_eltwise_layernorm_fuse_pass前一个
|
|
||
| auto axis = match_ctx.Attr<std::vector<int64_t>>("full_13_value"); | ||
| auto axis_2 = match_ctx.Attr<std::vector<int64_t>>("full_12_value"); | ||
| return check_axes(axis) && check_axes(axis_2); |
There was a problem hiding this comment.
这个地方return了，后面的check就不会执行了吧
Comment on lines
237
to
240
| return check_unsqueeze_axes(unsqueeze_axis) && | ||
| check_unsqueeze_axes(unsqueeze_axis_1) && | ||
| check_unsqueeze_axes(unsqueeze_axis_2) && | ||
| check_unsqueeze_axes(unsqueeze_axis_3); |
yuanlehome
approved these changes
Jun 20, 2024
co63oc
pushed a commit
to co63oc/Paddle
that referenced
this pull request
Jun 25, 2024
…#65265) * fused * 前面retrun true后面就不检查了 * 修改return true的后面无法检查 * 删除多余的注释
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Outdated suggestions cannot be applied.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
Pcard-71500
增加fused_rotary_position_embedding_pass
A30测试结果:
Llama-7b batch-size:1 no fused_rotary_position_embedding_pass:32857.339ms
Llama-7b batch-size:1 have fused_rotary_position_embedding_pass:31191.201ms
Llama-7b batch-size:4 no fused_rotary_position_embedding_pass:52057.086ms
Llama-7b batch-size:4 have fused_rotary_position_embedding_pass:50541.888ms
在pir 下，llama-7b 模型在 batch_size=1 与batch_size=4 的平均时延方面分别有 5.071% 与 2.911% 的性能提升，其余模型无明显影响；显存方面基本持平