Skip to content

P
Paddle

PaddlePaddle / Paddle
大约 1 年 前同步成功

通知 2299
Star 20931
Fork 5422
P
Paddle
未验证 提交 5c5536cb 编写于 作者: R RedContritio 提交者： GitHub
浏览文件
操作

Fix div 0 error of case25: paddle.dot (#50014) 

* support size 0 dot input

* prevent div 0 in grad

* add unittest

* remove unnecessary vlog

* add unittests
上级 aaaffd2e
隐藏空白更改
内联 并排
Showing with 92 addition and 30 deletion
paddle/phi/kernels/cpu/dot_kernel.cc
浏览文件 @ 5c5536cb
...@@ -35,7 +35,13 @@ void DotKernel(const Context& dev_ctx, ...@@ -35,7 +35,13 @@ void DotKernel(const Context& dev_ctx,
// B pairs along the way where B is the dimension of the least ordered axis // B pairs along the way where B is the dimension of the least ordered axis
auto&& d = x.dims(); auto&& d = x.dims();
auto const N = x.numel(); auto const N = x.numel();
auto const B = d[d.size() - 1];
// prevent div 0
auto const _B = d.size() == 0 ? 1 : d[d.size() - 1];
auto const B = _B != 0 ? _B : 1;
// initialize for N / B <= 0
z[0] = 0;
for (int j = 0; j < N / B; j++) { for (int j = 0; j < N / B; j++) {
T ss = 0; T ss = 0;
......
paddle/phi/kernels/impl/dot_grad_kernel_impl.h
浏览文件 @ 5c5536cb
...@@ -42,6 +42,7 @@ struct DotGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -42,6 +42,7 @@ struct DotGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
const DenseTensor* tensor_dout, const DenseTensor* tensor_dout,
DenseTensor* tensor_dx, DenseTensor* tensor_dx,
DenseTensor* tensor_dy) { DenseTensor* tensor_dy) {
VLOG(1) << "enable route";
#if defined(__NVCC__) || defined(__HIPCC__) #if defined(__NVCC__) || defined(__HIPCC__)
if (1 == tensor_dout->dims().size()) { if (1 == tensor_dout->dims().size()) {
auto dout = EigenVector<T>::Flatten(*tensor_dout); auto dout = EigenVector<T>::Flatten(*tensor_dout);
...@@ -103,7 +104,8 @@ struct DotGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -103,7 +104,8 @@ struct DotGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
const DDim& dim = tensor_x->dims(); const DDim& dim = tensor_x->dims();
size_t N = static_cast<size_t>(phi::product(dim)); size_t N = static_cast<size_t>(phi::product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -118,7 +120,8 @@ struct DotGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -118,7 +120,8 @@ struct DotGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
const DDim& dim = tensor_y->dims(); const DDim& dim = tensor_y->dims();
size_t N = static_cast<size_t>(phi::product(dim)); size_t N = static_cast<size_t>(phi::product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -182,7 +185,8 @@ struct DotGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -182,7 +185,8 @@ struct DotGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
*dz = tensor_dout->data<T>(); *dz = tensor_dout->data<T>();
auto&& d = tensor_x->dims(); auto&& d = tensor_x->dims();
auto const N = tensor_x->numel(); auto const N = tensor_x->numel();
auto const B = d[d.size() - 1]; auto const _B = d.size() == 0 ? 1 : d[d.size() - 1];
auto const B = _B != 0 ? _B : 1;
if (tensor_dx) { if (tensor_dx) {
auto* dx = ctx.template Alloc<T>(tensor_dx); auto* dx = ctx.template Alloc<T>(tensor_dx);
...@@ -298,7 +302,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -298,7 +302,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
const DDim& dim = tensor_dx->dims(); const DDim& dim = tensor_dx->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -316,7 +321,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -316,7 +321,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
const DDim& dim = tensor_dy->dims(); const DDim& dim = tensor_dy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -337,7 +343,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -337,7 +343,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
const DDim& dim = tensor_dy->dims(); const DDim& dim = tensor_dy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
...@@ -362,7 +369,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -362,7 +369,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
const DDim& dim = tensor_dy->dims(); const DDim& dim = tensor_dy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
...@@ -385,7 +393,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -385,7 +393,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
const DDim& dim = tensor_dx->dims(); const DDim& dim = tensor_dx->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
...@@ -475,7 +484,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -475,7 +484,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_dx = ctx.template Alloc<T>(tensor_dx); auto* data_dx = ctx.template Alloc<T>(tensor_dx);
const DDim& dim = tensor_dx->dims(); const DDim& dim = tensor_dx->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
if (0 == i % step) ++s; if (0 == i % step) ++s;
...@@ -490,7 +500,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -490,7 +500,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_dy = ctx.template Alloc<T>(tensor_dy); auto* data_dy = ctx.template Alloc<T>(tensor_dy);
const DDim& dim = tensor_dy->dims(); const DDim& dim = tensor_dy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
if (0 == i % step) ++s; if (0 == i % step) ++s;
...@@ -505,7 +516,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -505,7 +516,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_ddout = ctx.template Alloc<T>(tensor_ddout); auto* data_ddout = ctx.template Alloc<T>(tensor_ddout);
const DDim& dim = tensor_dy->dims(); const DDim& dim = tensor_dy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -524,7 +536,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -524,7 +536,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_ddout = ctx.template Alloc<T>(tensor_ddout); auto* data_ddout = ctx.template Alloc<T>(tensor_ddout);
const DDim& dim = tensor_dy->dims(); const DDim& dim = tensor_dy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -543,7 +556,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -543,7 +556,8 @@ struct DotDoubleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_ddout = ctx.template Alloc<T>(tensor_ddout); auto* data_ddout = ctx.template Alloc<T>(tensor_ddout);
const DDim& dim = tensor_dx->dims(); const DDim& dim = tensor_dx->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -771,7 +785,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -771,7 +785,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
auto* data_d_y = ctx.template Alloc<T>(out_tensor_d_y); auto* data_d_y = ctx.template Alloc<T>(out_tensor_d_y);
const DDim& dim = out_tensor_d_y->dims(); const DDim& dim = out_tensor_d_y->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -785,7 +800,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -785,7 +800,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
auto* data_d_dout = ctx.template Alloc<T>(out_tensor_d_dout); auto* data_d_dout = ctx.template Alloc<T>(out_tensor_d_dout);
const DDim& dim = in_tensor_x->dims(); const DDim& dim = in_tensor_x->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -811,7 +827,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -811,7 +827,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
auto* data_d_x = ctx.template Alloc<T>(out_tensor_d_x); auto* data_d_x = ctx.template Alloc<T>(out_tensor_d_x);
const DDim& dim = out_tensor_d_x->dims(); const DDim& dim = out_tensor_d_x->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -824,7 +841,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -824,7 +841,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
auto* data_d_dout = ctx.template Alloc<T>(out_tensor_d_dout); auto* data_d_dout = ctx.template Alloc<T>(out_tensor_d_dout);
const DDim& dim = in_tensor_x->dims(); const DDim& dim = in_tensor_x->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
if (d_dout_flag) { if (d_dout_flag) {
...@@ -859,7 +877,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -859,7 +877,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
auto* data_d_ddy = ctx.template Alloc<T>(out_tensor_d_ddy); auto* data_d_ddy = ctx.template Alloc<T>(out_tensor_d_ddy);
const DDim& dim = out_tensor_d_ddy->dims(); const DDim& dim = out_tensor_d_ddy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
if (0 == i % step) ++s; if (0 == i % step) ++s;
...@@ -875,7 +894,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -875,7 +894,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
auto* data_d_ddx = ctx.template Alloc<T>(out_tensor_d_ddx); auto* data_d_ddx = ctx.template Alloc<T>(out_tensor_d_ddx);
const DDim& dim = out_tensor_d_ddx->dims(); const DDim& dim = out_tensor_d_ddx->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
if (0 == i % step) ++s; if (0 == i % step) ++s;
...@@ -891,7 +911,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -891,7 +911,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
auto* data_d_ddx = ctx.template Alloc<T>(out_tensor_d_ddx); auto* data_d_ddx = ctx.template Alloc<T>(out_tensor_d_ddx);
const DDim& dim = out_tensor_d_ddx->dims(); const DDim& dim = out_tensor_d_ddx->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
if (d_ddx_flag) { if (d_ddx_flag) {
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -911,7 +932,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> { ...@@ -911,7 +932,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::EnableComplex<T>> {
auto* data_d_ddy = ctx.template Alloc<T>(out_tensor_d_ddy); auto* data_d_ddy = ctx.template Alloc<T>(out_tensor_d_ddy);
const DDim& dim = out_tensor_d_ddy->dims(); const DDim& dim = out_tensor_d_ddy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
if (d_ddy_flag) { if (d_ddy_flag) {
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -1144,7 +1166,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -1144,7 +1166,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_d_y = ctx.template Alloc<T>(out_tensor_d_y); auto* data_d_y = ctx.template Alloc<T>(out_tensor_d_y);
const DDim& dim = out_tensor_d_y->dims(); const DDim& dim = out_tensor_d_y->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
if (0 == i % step) ++s; if (0 == i % step) ++s;
...@@ -1155,7 +1178,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -1155,7 +1178,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_d_dout = ctx.template Alloc<T>(out_tensor_d_dout); auto* data_d_dout = ctx.template Alloc<T>(out_tensor_d_dout);
const DDim& dim = in_tensor_x->dims(); const DDim& dim = in_tensor_x->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -1179,7 +1203,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -1179,7 +1203,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_d_x = ctx.template Alloc<T>(out_tensor_d_x); auto* data_d_x = ctx.template Alloc<T>(out_tensor_d_x);
const DDim& dim = out_tensor_d_x->dims(); const DDim& dim = out_tensor_d_x->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
if (0 == i % step) ++s; if (0 == i % step) ++s;
...@@ -1190,7 +1215,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -1190,7 +1215,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_d_dout = ctx.template Alloc<T>(out_tensor_d_dout); auto* data_d_dout = ctx.template Alloc<T>(out_tensor_d_dout);
const DDim& dim = in_tensor_x->dims(); const DDim& dim = in_tensor_x->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
bool new_s = false; bool new_s = false;
if (d_dout_flag) { if (d_dout_flag) {
...@@ -1222,7 +1248,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -1222,7 +1248,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_d_ddy = ctx.template Alloc<T>(out_tensor_d_ddy); auto* data_d_ddy = ctx.template Alloc<T>(out_tensor_d_ddy);
const DDim& dim = out_tensor_d_ddy->dims(); const DDim& dim = out_tensor_d_ddy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
if (0 == i % step) ++s; if (0 == i % step) ++s;
...@@ -1237,7 +1264,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -1237,7 +1264,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_d_ddx = ctx.template Alloc<T>(out_tensor_d_ddx); auto* data_d_ddx = ctx.template Alloc<T>(out_tensor_d_ddx);
const DDim& dim = out_tensor_d_ddx->dims(); const DDim& dim = out_tensor_d_ddx->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
if (0 == i % step) ++s; if (0 == i % step) ++s;
...@@ -1252,7 +1280,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -1252,7 +1280,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_d_ddx = ctx.template Alloc<T>(out_tensor_d_ddx); auto* data_d_ddx = ctx.template Alloc<T>(out_tensor_d_ddx);
const DDim& dim = out_tensor_d_ddx->dims(); const DDim& dim = out_tensor_d_ddx->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
if (d_ddx_flag) { if (d_ddx_flag) {
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
...@@ -1270,7 +1299,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> { ...@@ -1270,7 +1299,8 @@ struct DotTripleGradFunction<DeviceContext, T, phi::funcs::DisableComplex<T>> {
auto* data_d_ddy = ctx.template Alloc<T>(out_tensor_d_ddy); auto* data_d_ddy = ctx.template Alloc<T>(out_tensor_d_ddy);
const DDim& dim = out_tensor_d_ddy->dims(); const DDim& dim = out_tensor_d_ddy->dims();
size_t N = static_cast<size_t>(product(dim)); size_t N = static_cast<size_t>(product(dim));
auto step = dim[dim.size() - 1]; auto _step = dim.size() > 0 ? dim[dim.size() - 1] : 1;
auto step = _step != 0 ? _step : 1;
int s = -1; int s = -1;
if (d_ddy_flag) { if (d_ddy_flag) {
for (size_t i = 0; i < N; ++i) { for (size_t i = 0; i < N; ++i) {
......
python/paddle/fluid/tests/unittests/test_dot_op.py
浏览文件 @ 5c5536cb
...@@ -92,6 +92,32 @@ class DotOp(OpTest): ...@@ -92,6 +92,32 @@ class DotOp(OpTest):
self.dtype = np.float64 self.dtype = np.float64
class DotOpEmptyInput(unittest.TestCase):
def test_1d_input(self):
data = np.array([], dtype=np.float32)
x = paddle.to_tensor(np.reshape(data, [0]), dtype='float32')
y = paddle.to_tensor(np.reshape(data, [0]), dtype='float32')
np_out = np.dot(data, data)
pd_out = paddle.dot(x, y)
self.assertEquals(np_out, pd_out)
def test_2d_input(self):
data = np.array([], dtype=np.float32)
x = paddle.to_tensor(np.reshape(data, [0, 0]), dtype='float32')
y = paddle.to_tensor(np.reshape(data, [0, 0]), dtype='float32')
pd_out = paddle.dot(x, y)
self.assertEqual(pd_out.shape, (0, 1))
def test_3d_input_error(self):
data = np.array([], dtype=np.float32)
x = paddle.to_tensor(np.reshape(data, [0, 0, 0]), dtype='float32')
y = paddle.to_tensor(np.reshape(data, [0, 0, 0]), dtype='float32')
self.assertRaises(Exception, paddle.dot, x, y)
class DotOpBatch(DotOp): class DotOpBatch(DotOp):
def init_input_output(self): def init_input_output(self):
self.x = ( self.x = (
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或