【Operator Mechanism】add fused_stack_transpose_quant api #73228
Merged
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! |
lshpku
changed the title
ggggxm
force-pushed
the
stack_tranpose_quant
branch
from
b8eefa1 to
daff73d
Compare
zhangbo9674
reviewed
Jun 11, 2025
paddle/phi/ops/yaml/ops.yaml
Outdated
| @@ -2382,6 +2382,24 @@ | |||
| backward: fused_softmax_mask_upper_triangle_grad | |||
| interfaces : paddle::dialect::InferSymbolicShapeInterface | |||
|
|
|||
| - op : fused_stack_quant | |||
There was a problem hiding this comment.
有一个专门存放融合算子的：paddle/phi/ops/yaml/fused_ops.yaml
| #include <iostream> | ||
| #include <limits> | ||
|
|
||
| // #include "paddle/extension.h" |
| } \ | ||
| } | ||
|
|
||
| // 对二维坐标进行swizzle变换，提供相对offset,避免bank conflict |
ggggxm
force-pushed
the
stack_tranpose_quant
branch
from
daff73d to
149c8c1
Compare
zyfncg
reviewed
Jun 13, 2025
paddle/phi/infermeta/multiary.cc
Outdated
| @@ -6245,6 +6245,108 @@ void FullWithTensorInferMeta(const IntArray& shape, | |||
| out->set_dtype(dtype); | |||
| } | |||
|
|
|||
| std::tuple<int64_t, int64_t, int64_t> FusedStackQuantCommonCheck( | |||
paddle/phi/infermeta/multiary.h
Outdated
Comment on lines
553
to
554
| std::tuple<int64_t, int64_t, int64_t> FusedStackQuantCommonCheck( | ||
| const std::vector<const MetaTensor*>& x); |
| @@ -0,0 +1,84 @@ | |||
| # Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved. | |||
python/paddle/nn/functional/fp8.py
Outdated
|
|
||
|
|
||
| """ | ||
| if in_dynamic_mode(): |
python/paddle/nn/functional/fp8.py
Outdated
| else: | ||
| return _C_ops.fused_stack_quant(x) | ||
|
|
||
| else: |
lshpku
reviewed
Jun 13, 2025
Comment on lines
79
to
84
| dev_ctx.template Alloc<phi::dtype::float8_e4m3fn>(out); | ||
| dev_ctx.template Alloc<float>(scale); | ||
| auto out_dims = out->dims(); | ||
| out->Resize(out_dims); | ||
| auto scale_dims = scale->dims(); | ||
| scale->Resize(scale_dims); |
There was a problem hiding this comment.
一般是先resize才alloc：
Paddle/paddle/fluid/framework/operator.h
Lines 594 to 597 in c999558
还有为啥要把一个tensor resize成自己的dims？这样会产生什么变化吗
ggggxm
force-pushed
the
stack_tranpose_quant
branch
from
f6798e5 to
4ddde16
Compare
zyfncg
previously approved these changes
Jun 13, 2025
This reverts commit 06ca104.
lshpku
approved these changes
Jun 14, 2025
zhangbo9674
approved these changes
Jun 14, 2025
zyfncg
approved these changes
Jun 14, 2025
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
Operator Mechanism
PR Types
New features
Description
新增fused_stack_transpose_quant和fused_stack_quant支持