add masked fill op #3515
Merged
add masked fill op #3515
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
liujuncheng
reviewed
Aug 28, 2020
oneflow/user/ops/masked_fill_op.cpp
Outdated
| REGISTER_USER_OP("masked_fill") | ||
| .Input("x") | ||
| .Input("mask") | ||
| .Attr("value", UserOpAttrType::kAtFloat) |
There was a problem hiding this comment.
masked_fill支持多种数据类型，没有办法用float精确表达，这里可以参考scalar_mul
liujuncheng
reviewed
Aug 31, 2020
added 8 commits
August 31, 2020 17:09
…dev_add_op_masked_fill
…dev_add_op_masked_fill
liujuncheng
reviewed
Sep 2, 2020
|
|
||
| namespace { | ||
|
|
||
| __global__ void NaiveHalfFillGpu(const int64_t elem_cnt, const float16 x, float16* y) { |
There was a problem hiding this comment.
NewKernelUtil<device_type>::Fill是支持fp16的，是不是只需要处理好operand到float16的转换，而不是新写kernel
|
|
||
| REGISTER_HALF_CONSTANT_LIKE_KERNEL | ||
|
|
||
| } // namespace oneflow |
|
|
||
| namespace { | ||
|
|
||
| __global__ void HalfAddByScalarPtrGpu(const int64_t n, const half* x, const half* y, half* z) { |
There was a problem hiding this comment.
这里应该是给XxxByScalarPtr添加float16类型吧
| namespace { | ||
|
|
||
| template<typename CondT> | ||
| __global__ void NaiveHalfWhere(const int64_t elem_cnt, const CondT* cond, const half* lhs, |
There was a problem hiding this comment.
同理，这里直接给where添加float16就可以了吧
|
|
||
| __global__ void HalfAddByScalarPtrGpu(const int64_t n, const half* x, const half* y, half* z) { | ||
| const half y_value = y[0]; | ||
| CUDA_1D_KERNEL_LOOP(i, n) { z[i] = x[i] + y_value; } |
There was a problem hiding this comment.
float16/half类型不能直接用+，要用__hadd
change piar to macro Co-authored-by: Juncheng <liujuncheng1022@gmail.com>
liujuncheng
approved these changes
Sep 2, 2020
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.
添加masked fill op
reference https://pytorch.org/docs/stable/tensors.html?highlight=masked_fill#torch.Tensor.masked_fill