mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 12:21:27 +01:00
9243b64bff
* Update elementwise ops to support numpy style boradcast Update elementwise ops to support numpy style boradcast * Fix sqrt_op * Fix compare ops * Fix gradient test * Fix optimizer legacy broadcast * Fix legacy broadcast for elementwise ops * Skip flaky test * Fix eigen simple binary op * Fix attention test * Fix rnn test * Fix LSTM test * Fix tan grad * Fix schema check
49 lines
1.0 KiB
C++
49 lines
1.0 KiB
C++
#ifndef CAFFE2_OPERATORS_ELEMENTWISE_DIV_OP_H_
|
|
#define CAFFE2_OPERATORS_ELEMENTWISE_DIV_OP_H_
|
|
|
|
#include <vector>
|
|
|
|
#include "caffe2/operators/elementwise_ops.h"
|
|
#include "caffe2/utils/math.h"
|
|
|
|
namespace caffe2 {
|
|
|
|
template <class Context>
|
|
struct DivFunctor {
|
|
template <typename TIn, typename TOut>
|
|
bool Forward(
|
|
const std::vector<int>& A_dims,
|
|
const std::vector<int>& B_dims,
|
|
const TIn* A,
|
|
const TIn* B,
|
|
TOut* C,
|
|
Context* context) const {
|
|
math::Div(
|
|
A_dims.size(),
|
|
A_dims.data(),
|
|
B_dims.size(),
|
|
B_dims.data(),
|
|
A,
|
|
B,
|
|
C,
|
|
context);
|
|
return true;
|
|
}
|
|
|
|
template <typename TGrad, typename TIn, typename TOut>
|
|
bool Backward(
|
|
const std::vector<int>& A_dims,
|
|
const std::vector<int>& B_dims,
|
|
const TGrad* dC_data,
|
|
const TIn* A_data,
|
|
const TIn* B_data,
|
|
const TOut* C_data,
|
|
TGrad* dA_data,
|
|
TGrad* dB_data,
|
|
Context* context) const;
|
|
};
|
|
|
|
} // namespace caffe2
|
|
|
|
#endif // CAFFE2_OPERATORS_ELEMENTWISE_DIV_OP_H_
|