Add is_mean param for mean op #40757
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! |
ZzSean
reviewed
Mar 21, 2022
| @@ -65,9 +65,10 @@ class MeanCUDAKernel : public framework::OpKernel<T> { | |||
| for (decltype(rank) i = 0; i < rank; ++i) { | |||
| reduce_dims.push_back(i); | |||
| } | |||
| TensorReduceImpl<T, T, kernel_primitives::AddFunctor, Div>( | |||
| context.cuda_device_context(), *input, output, Div(numel), reduce_dims, | |||
There was a problem hiding this comment.
已经删除了调用的是IdentityFunctor
ZzSean
reviewed
Mar 21, 2022
| @@ -657,6 +658,9 @@ __global__ void ReduceAnyKernel(const Tx* x, | |||
| // the last dim gets involved in reduction | |||
| int store_offset = 0; | |||
| int stride_left = 0; | |||
| auto Final = | |||
| is_mean ? kps::DivideFunctor<MPType> : kps::IdentityFunctor<MPType>; | |||
| auto final_opt = Final(reduce_num); | |||
There was a problem hiding this comment.
opt一般指的是optimize的缩写吧，如果这里含义是output建议直接用out或者output
There was a problem hiding this comment.
不是out，是operat，进行除法操作，或者是直接返回，只针对最后的store数据进行操作
ZzSean
reviewed
Mar 21, 2022
| @@ -657,6 +658,9 @@ __global__ void ReduceAnyKernel(const Tx* x, | |||
| // the last dim gets involved in reduction | |||
| int store_offset = 0; | |||
| int stride_left = 0; | |||
| auto Final = | |||
| is_mean ? kps::DivideFunctor<MPType> : kps::IdentityFunctor<MPType>; | |||
There was a problem hiding this comment.
对于is_mean为false的情况，感觉不用再用IdentityFunctor算一遍了，只在最后把reduce_var除一下就可以
zkh2016
reviewed
Mar 24, 2022
| @@ -33,12 +33,12 @@ void TensorReduceImpl(const platform::CUDADeviceContext& dev_ctx, | |||
| const framework::Tensor& x, framework::Tensor* y, | |||
| const TransformOp& transform, | |||
| const std::vector<int>& origin_reduce_dims, | |||
| gpuStream_t stream) { | |||
| gpuStream_t stream, bool is_mean = false) { | |||
zkh2016
approved these changes
Mar 24, 2022
xingfeng01
approved these changes
Mar 24, 2022
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
OPs
Describe
Add is_mean param for mean op
1.针对mean OP 添加is_mean参数，保证完成所以数据求和之后再进行除法操作。
2.修改reduceHigher的grid配置，当grid.z > 65536时候设置reduce_type为reduceAny
修改背景，1. fp16状态下，模型出现nan， 2， 模型case[6600,:,:] 计算错误 axis = 1