【PIR】add sequence_mask in pir #59348
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! |
kangguangli
reviewed
Nov 27, 2023
kangguangli
previously approved these changes
Dec 4, 2023
YuanRisheng
reviewed
Dec 4, 2023
| args: (Tensor x, Scalar(int) max_len, int out_dtype) | ||
| output: Tensor(y) | ||
| infer_meta: | ||
| func: SequenceMaskPIRInferMeta |
There was a problem hiding this comment.
phi算子库下的算子和具体执行体系无关，后续也并非pir专用的，这里命名不太合理，建议改成SequenceMaskScalarInferMeta
| infer_meta: | ||
| func: SequenceMaskPIRInferMeta | ||
| kernel: | ||
| func: sequence_mask_pir |
There was a problem hiding this comment.
同上，建议改成sequence_mask_scalar
Comment on lines
95
to
124
| int maxlen = max_len.to<int>(); | ||
| auto* x_data = x.data<T>(); | ||
| auto x_numel = x.numel(); | ||
|
|
||
| if (maxlen < 0) { | ||
| if (x_numel == 0) { | ||
| maxlen = 0; | ||
| } else { | ||
| #if defined(__NVCC__) || defined(__HIPCC__) | ||
| VLOG(10) | ||
| << "SequenceMaskOp on GPU may be slow when maxlen is not provided."; | ||
| maxlen = static_cast<int>( | ||
| thrust::reduce(thrust::device_pointer_cast(x_data), | ||
| thrust::device_pointer_cast(x_data) + x_numel, | ||
| static_cast<T>(0), | ||
| thrust::maximum<T>())); | ||
| #else | ||
| maxlen = static_cast<int>(*std::max_element(x_data, x_data + x_numel)); | ||
| #endif | ||
| } | ||
| } | ||
|
|
||
| auto y_dim = phi::vectorize<int>(x.dims()); | ||
| y_dim.push_back(maxlen); | ||
| y->Resize(phi::make_ddim(y_dim)); | ||
|
|
||
| phi::VisitDataType(phi::TransToPhiDataType(out_dtype), | ||
| phi::funcs::SequenceMaskFunctor<Context, T>( | ||
| ctx, x_data, y, x_numel * maxlen, maxlen)); | ||
| } |
There was a problem hiding this comment.
既然这里已经将代码拷贝过来了，可以将SequenceMaskKernel中的重复代码删掉，通过传入的max_len_tensor或者maxlen，构造一个Scalar类型的max_len，调用SequenceMaskScalarKernel即可
yangguohao
force-pushed
the
sequence_mask
branch
from
cd19ed2
to
c19fa9b
Compare
yangguohao
force-pushed
the
sequence_mask
branch
from
c19fa9b
to
a3e7e2c
Compare
yangguohao
force-pushed
the
sequence_mask
branch
from
f781324
to
0d378ab
Compare
kangguangli
previously approved these changes
Dec 7, 2023
YuanRisheng
previously approved these changes
Dec 7, 2023
jeff41404
previously approved these changes
Dec 7, 2023
zyfncg
previously approved these changes
Dec 7, 2023
heavyrain-lzy
previously approved these changes
Dec 7, 2023
yangguohao
dismissed stale reviews from heavyrain-lzy, zyfncg, jzhang533, jeff41404, XieYunshen, YuanRisheng, and kangguangli
via
b6300af
kangguangli
approved these changes
Dec 8, 2023
XieYunshen
approved these changes
Dec 8, 2023
jeff41404
approved these changes
Dec 8, 2023
zyfncg
approved these changes
Dec 8, 2023
20 tasks
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 types
Others
PR changes
Others
Description
add sequence_mask for pir