Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

【Hackathon + No.60】 angle 等算子 FP16/BF16算子及单测完善 #51666

Closed
wants to merge 8 commits into from
Closed

【Hackathon + No.60】 angle 等算子 FP16/BF16算子及单测完善 #51666

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
fa50e76
all
yuhao-su Mar 13, 2023
7e717bf
Add files via upload
KateJing1212 Mar 14, 2023
6406055
Add files via upload
KateJing1212 Mar 14, 2023
28fe4bc
fix
yuhao-su Mar 14, 2023
1f10e82
Merge pull request #1 from KateJing1212/KateJing1212-patch-1
yuhao-su Mar 14, 2023
8a0a24f
conflict
yuhao-su Mar 14, 2023
e77214a
conflict
yuhao-su Mar 15, 2023
03ad3ce
codestyle
KateJing1212 Mar 16, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
341 changes: 203 additions & 138 deletions paddle/fluid/distributed/ps/thirdparty/round_robin.h
View file
Open in desktop

Large diffs are not rendered by default.

18 changes: 10 additions & 8 deletions paddle/fluid/inference/tensorrt/plugin/transformer_input_output_convert_plugin.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,10 +44,11 @@ class TransformerInputConvertPlugin : public DynamicPluginTensorRT {

int initialize() TRT_NOEXCEPT { return 0; }
void terminate() TRT_NOEXCEPT;
nvinfer1::DimsExprs getOutputDimensions(int outputIndex,
const nvinfer1::DimsExprs* inputs,
int nbInputs,
nvinfer1::IExprBuilder& exprBuilder) // NOLINT
nvinfer1::DimsExprs getOutputDimensions(
int outputIndex,
const nvinfer1::DimsExprs* inputs,
int nbInputs,
nvinfer1::IExprBuilder& exprBuilder) // NOLINT
TRT_NOEXCEPT override;

bool supportsFormatCombination(int pos,
Expand Down Expand Up @@ -155,10 +156,11 @@ class TransformerOutputConvertPlugin : public DynamicPluginTensorRT {

int initialize() TRT_NOEXCEPT { return 0; }
void terminate() TRT_NOEXCEPT;
nvinfer1::DimsExprs getOutputDimensions(int outputIndex,
const nvinfer1::DimsExprs* inputs,
int nbInputs,
nvinfer1::IExprBuilder& exprBuilder) // NOLINT
nvinfer1::DimsExprs getOutputDimensions(
int outputIndex,
const nvinfer1::DimsExprs* inputs,
int nbInputs,
nvinfer1::IExprBuilder& exprBuilder) // NOLINT
TRT_NOEXCEPT override;

bool supportsFormatCombination(int pos,
Expand Down
1 change: 1 addition & 0 deletions paddle/phi/kernels/clip_by_norm_kernel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@

#pragma once

#include "paddle/phi/common/scalar.h"
#include "paddle/phi/core/dense_tensor.h"

namespace phi {
Expand Down
9 changes: 7 additions & 2 deletions paddle/phi/kernels/cpu/clip_by_norm_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,5 +30,10 @@ void ClipByNormKernel(const Context& dev_ctx,

} // namespace phi

PD_REGISTER_KERNEL(
clip_by_norm, CPU, ALL_LAYOUT, phi::ClipByNormKernel, float) {}
PD_REGISTER_KERNEL(clip_by_norm,
CPU,
ALL_LAYOUT,
phi::ClipByNormKernel,
float,
phi::dtype::float16,
phi::dtype::bfloat16) {}
10 changes: 8 additions & 2 deletions paddle/phi/kernels/cpu/multi_dot_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,5 +18,11 @@ limitations under the License. */
#include "paddle/phi/core/kernel_registry.h"
#include "paddle/phi/kernels/impl/multi_dot_kernel_impl.h"

PD_REGISTER_KERNEL(
multi_dot, CPU, ALL_LAYOUT, phi::MultiDotKernel, float, double) {}
PD_REGISTER_KERNEL(multi_dot,
CPU,
ALL_LAYOUT,
phi::MultiDotKernel,
float,
double,
phi::dtype::float16,
phi::dtype::bfloat16) {}
10 changes: 8 additions & 2 deletions paddle/phi/kernels/cpu/multinomial_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,5 +44,11 @@ void MultinomialKernel(const Context& dev_ctx,

} // namespace phi

PD_REGISTER_KERNEL(
multinomial, CPU, ALL_LAYOUT, phi::MultinomialKernel, float, double) {}
PD_REGISTER_KERNEL(multinomial,
CPU,
ALL_LAYOUT,
phi::MultinomialKernel,
float,
double,
phi::dtype::float16,
phi::dtype::bfloat16) {}
4 changes: 3 additions & 1 deletion paddle/phi/kernels/cpu/overlap_add_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -147,4 +147,6 @@ PD_REGISTER_KERNEL(overlap_add,
float,
double,
phi::dtype::complex<float>,
phi::dtype::complex<double>) {}
phi::dtype::complex<double>,
phi::dtype::float16,
phi::dtype::bfloat16) {}
9 changes: 8 additions & 1 deletion paddle/phi/kernels/cpu/prelu_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,4 +68,11 @@ void PReluKernel(const Context& dev_ctx,

} // namespace phi

PD_REGISTER_KERNEL(prelu, CPU, ALL_LAYOUT, phi::PReluKernel, float, double) {}
PD_REGISTER_KERNEL(prelu,
CPU,
ALL_LAYOUT,
phi::PReluKernel,
float,
double,
phi::dtype::float16,
phi::dtype::bfloat16) {}
15 changes: 14 additions & 1 deletion paddle/phi/kernels/cpu/randperm_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,11 +42,24 @@ void RandpermKernel(const Context& dev_ctx,

} // namespace phi

PD_REGISTER_KERNEL(randperm_raw,
CPU,
ALL_LAYOUT,
phi::RandpermRawKernel,
float,
double,
int,
int64_t,
phi::dtype::float16,
phi::dtype::bfloat16) {}

PD_REGISTER_KERNEL(randperm,
CPU,
ALL_LAYOUT,
phi::RandpermKernel,
float,
double,
int,
int64_t) {}
int64_t,
phi::dtype::float16,
phi::dtype::bfloat16) {}
12 changes: 8 additions & 4 deletions paddle/phi/kernels/cpu/sign_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,11 @@ limitations under the License. */
#include "paddle/phi/core/kernel_registry.h"
#include "paddle/phi/kernels/impl/sign_kernel_impl.h"

// See Note [ Why still include the fluid headers? ]
#include "paddle/phi/common/bfloat16.h"

PD_REGISTER_KERNEL(sign, CPU, ALL_LAYOUT, phi::SignKernel, float, double) {}
PD_REGISTER_KERNEL(sign,
CPU,
ALL_LAYOUT,
phi::SignKernel,
float,
double,
phi::dtype::float16,
phi::dtype::bfloat16) {}
3 changes: 2 additions & 1 deletion paddle/phi/kernels/gpu/clip_by_norm_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,4 +86,5 @@ PD_REGISTER_KERNEL(clip_by_norm,
ALL_LAYOUT,
phi::ClipByNormKernel,
float,
phi::dtype::float16) {}
phi::dtype::float16,
phi::dtype::bfloat16) {}
10 changes: 8 additions & 2 deletions paddle/phi/kernels/gpu/multi_dot_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,5 +21,11 @@ limitations under the License. */

using float16 = phi::dtype::float16;

PD_REGISTER_KERNEL(
multi_dot, GPU, ALL_LAYOUT, phi::MultiDotKernel, float, double, float16) {}
PD_REGISTER_KERNEL(multi_dot,
GPU,
ALL_LAYOUT,
phi::MultiDotKernel,
float,
double,
float16,
phi::dtype::bfloat16) {}
4 changes: 3 additions & 1 deletion paddle/phi/kernels/gpu/multinomial_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -296,6 +296,8 @@ PD_REGISTER_KERNEL(multinomial, // cuda_only
ALL_LAYOUT,
phi::MultinomialKernel,
float,
double) {}
double,
phi::dtype::float16,
phi::dtype::bfloat16) {}

#endif
5 changes: 3 additions & 2 deletions paddle/phi/kernels/gpu/overlap_add_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,6 +146,7 @@ PD_REGISTER_KERNEL(overlap_add,
int64_t,
float,
double,
phi::dtype::float16,
phi::dtype::complex<float>,
phi::dtype::complex<double>) {}
phi::dtype::complex<double>,
phi::dtype::float16,
phi::dtype::bfloat16) {}
1 change: 1 addition & 0 deletions paddle/phi/kernels/gpu/prelu_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,4 +68,5 @@ PD_REGISTER_KERNEL(prelu,
phi::PReluKernel,
float,
phi::dtype::float16,
phi::dtype::bfloat16,
double) {}
15 changes: 14 additions & 1 deletion paddle/phi/kernels/gpu/randperm_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -156,11 +156,24 @@ void RandpermKernel(const Context& dev_ctx,

} // namespace phi

PD_REGISTER_KERNEL(randperm_raw,
GPU,
ALL_LAYOUT,
phi::RandpermRawKernel,
float,
double,
int,
int64_t,
phi::dtype::float16,
phi::dtype::bfloat16) {}

PD_REGISTER_KERNEL(randperm,
GPU,
ALL_LAYOUT,
phi::RandpermKernel,
float,
double,
int,
int64_t) {}
int64_t,
phi::dtype::float16,
phi::dtype::bfloat16) {}
10 changes: 8 additions & 2 deletions paddle/phi/kernels/gpu/sign_kernel.cu.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,5 +23,11 @@ limitations under the License. */

using float16 = phi::dtype::float16;

PD_REGISTER_KERNEL(
sign, GPU, ALL_LAYOUT, phi::SignKernel, float, double, float16) {}
PD_REGISTER_KERNEL(sign,
GPU,
ALL_LAYOUT,
phi::SignKernel,
float,
double,
float16,
phi::dtype::bfloat16) {}
1 change: 1 addition & 0 deletions paddle/phi/kernels/multi_dot_kernel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@

#pragma once

#include "paddle/phi/common/scalar.h"
#include "paddle/phi/core/dense_tensor.h"

namespace phi {
Expand Down
1 change: 1 addition & 0 deletions paddle/phi/kernels/overlap_add_kernel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@

#pragma once

#include "paddle/phi/common/scalar.h"
#include "paddle/phi/core/dense_tensor.h"

namespace phi {
Expand Down
1 change: 1 addition & 0 deletions paddle/phi/kernels/prelu_kernel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@

#pragma once

#include "paddle/phi/common/scalar.h"
#include "paddle/phi/core/dense_tensor.h"

namespace phi {
Expand Down
1 change: 1 addition & 0 deletions paddle/phi/kernels/randperm_kernel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@

#pragma once

#include "paddle/phi/common/scalar.h"
#include "paddle/phi/core/dense_tensor.h"
#include "paddle/phi/core/device_context.h"

Expand Down
1 change: 1 addition & 0 deletions paddle/phi/kernels/sign_kernel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@ limitations under the License. */

#pragma once

#include "paddle/phi/common/scalar.h"
#include "paddle/phi/core/dense_tensor.h"
#include "paddle/phi/infermeta/unary.h"
#include "paddle/phi/kernels/empty_kernel.h"
Expand Down
Loading