Skip to content

[XPU] bfloat16 support for gather/gather_grad/scatter/scatter_grad #71132

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

Merged
merged 1 commit into from
Feb 17, 2025
Merged

[XPU] bfloat16 support for gather/gather_grad/scatter/scatter_grad #71132

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion cmake/external/xpu.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ set(XPU_XFA_LIB_NAME "libxpu_flash_attention.so")
set(XPU_XPUDNN_LIB_NAME "libxpu_dnn.so")

if(NOT DEFINED XPU_XHPC_BASE_DATE)
set(XPU_XHPC_BASE_DATE "dev/20250204")
set(XPU_XHPC_BASE_DATE "dev/20250213")
endif()
set(XPU_XCCL_BASE_VERSION "3.0.2.1") # For XRE5
if(NOT DEFINED XPU_XFT_BASE_VERSION)
Expand Down
18 changes: 16 additions & 2 deletions paddle/phi/backends/xpu/xpu3_op_list.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -701,7 +701,14 @@ XPUOpMap& get_kl3_ops() {
XPUKernelSet({phi::DataType::FLOAT32, phi::DataType::FLOAT16})},
{"floor", XPUKernelSet({phi::DataType::FLOAT32})},
{"gather_grad",
XPUKernelSet({phi::DataType::FLOAT32, phi::DataType::FLOAT16})},
XPUKernelSet({phi::DataType::FLOAT32,
phi::DataType::FLOAT16,
phi::DataType::BFLOAT16,
phi::DataType::INT8,
phi::DataType::INT16,
phi::DataType::INT32,
phi::DataType::INT64,
phi::DataType::BOOL})},
{"gather_nd_grad",
XPUKernelSet({phi::DataType::INT32,
phi::DataType::INT64,
Expand All @@ -717,6 +724,9 @@ XPUOpMap& get_kl3_ops() {
{"gather",
XPUKernelSet({phi::DataType::FLOAT32,
phi::DataType::FLOAT16,
phi::DataType::BFLOAT16,
phi::DataType::INT8,
phi::DataType::INT16,
phi::DataType::INT32,
phi::DataType::INT64,
phi::DataType::BOOL})},
Expand Down Expand Up @@ -1154,9 +1164,13 @@ XPUOpMap& get_kl3_ops() {
{"scatter",
XPUKernelSet({phi::DataType::INT64,
phi::DataType::INT32,
phi::DataType::FLOAT16,
phi::DataType::BFLOAT16,
phi::DataType::FLOAT32})},
{"scatter_grad",
XPUKernelSet({phi::DataType::FLOAT16, phi::DataType::FLOAT32})},
XPUKernelSet({phi::DataType::FLOAT16,
phi::DataType::BFLOAT16,
phi::DataType::FLOAT32})},
{"scatter_nd_add",
XPUKernelSet({phi::DataType::FLOAT32,
phi::DataType::INT32,
Expand Down
5 changes: 4 additions & 1 deletion paddle/phi/kernels/cpu/gather_grad_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,9 +72,12 @@ PD_REGISTER_KERNEL(gather_grad,
phi::GatherGradKernel,
float,
double,
int,
uint8_t,
int8_t,
int16_t,
int32_t,
int64_t,
bool,
phi::dtype::bfloat16,
phi::dtype::complex<float>,
phi::dtype::complex<double>) {}
5 changes: 4 additions & 1 deletion paddle/phi/kernels/cpu/gather_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,9 +70,12 @@ PD_REGISTER_KERNEL(gather,
phi::GatherKernel,
float,
double,
int,
uint8_t,
int8_t,
int16_t,
int32_t,
int64_t,
bool,
phi::dtype::bfloat16,
phi::dtype::complex<float>,
phi::dtype::complex<double>) {}
6 changes: 6 additions & 0 deletions paddle/phi/kernels/fusion/xpu/fused_rope_utils.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -358,11 +358,13 @@ void XPUFusedRotaryHalf(const Context& dev_ctx,
nullptr,
reinterpret_cast<const XPUSCType*>(sin_data),
reinterpret_cast<const XPUSCType*>(cos_data),
nullptr,
reinterpret_cast<XPUType*>(out_q->data()),
nullptr,
{batch_size, seq_len, num_heads, head_dim},
{batch_size, seq_len, 1, head_dim},
{},
0,
"BLHD",
-1,
10000.0f);
Expand All @@ -374,11 +376,13 @@ void XPUFusedRotaryHalf(const Context& dev_ctx,
reinterpret_cast<const XPUType*>(in_k->data()),
reinterpret_cast<const XPUSCType*>(sin_data),
reinterpret_cast<const XPUSCType*>(cos_data),
nullptr,
reinterpret_cast<XPUType*>(out_q->data()),
reinterpret_cast<XPUType*>(out_k->data()),
{batch_size, seq_len, num_heads, head_dim},
{batch_size, seq_len, 1, head_dim},
{},
0,
"BLHD",
num_heads_k,
10000.0f);
Expand All @@ -392,11 +396,13 @@ void XPUFusedRotaryHalf(const Context& dev_ctx,
nullptr,
reinterpret_cast<const XPUSCType*>(sin_data),
reinterpret_cast<const XPUSCType*>(cos_data),
nullptr,
reinterpret_cast<XPUType*>(out_v->data()),
nullptr,
{batch_size, seq_len, num_heads_v, head_dim},
{batch_size, seq_len, 1, head_dim},
{},
0,
"BLHD",
-1,
10000.0f);
Expand Down
8 changes: 7 additions & 1 deletion paddle/phi/kernels/xpu/gather_grad_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -101,4 +101,10 @@ PD_REGISTER_KERNEL(gather_grad,
ALL_LAYOUT,
phi::GatherGradKernel,
float,
phi::dtype::float16) {}
phi::dtype::float16,
phi::dtype::bfloat16,
int8_t,
int16_t,
int32_t,
int64_t,
bool) {}
5 changes: 4 additions & 1 deletion paddle/phi/kernels/xpu/gather_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -88,6 +88,9 @@ PD_REGISTER_KERNEL(gather,
phi::GatherKernel,
float,
phi::dtype::float16,
int,
phi::dtype::bfloat16,
int8_t,
int16_t,
int32_t,
int64_t,
bool) {}
3 changes: 2 additions & 1 deletion paddle/phi/kernels/xpu/scatter_grad_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,4 +93,5 @@ PD_REGISTER_KERNEL(scatter_grad,
ALL_LAYOUT,
phi::ScatterGradKernel,
float,
phi::dtype::float16) {}
phi::dtype::float16,
phi::dtype::bfloat16) {}
5 changes: 3 additions & 2 deletions paddle/phi/kernels/xpu/scatter_kernel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,6 +119,7 @@ PD_REGISTER_KERNEL(scatter,
ALL_LAYOUT,
phi::ScatterKernel,
float,
int,
int32_t,
int64_t,
phi::dtype::float16) {}
phi::dtype::float16,
phi::dtype::bfloat16) {}
7 changes: 6 additions & 1 deletion test/xpu/test_gather_op_xpu.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,7 @@
create_test_class,
get_xpu_op_support_types,
)
from op_test import convert_float_to_uint16
from op_test_xpu import XPUOpTest

import paddle
Expand All @@ -45,7 +46,11 @@ def setUp(self):
self.dtype = self.in_type

self.init_config()
xnp = np.random.random(self.x_shape).astype(self.dtype)
xnp = np.random.random(self.x_shape).astype(
self.dtype if self.dtype != np.uint16 else np.float32
)
if self.dtype == np.uint16:
xnp = convert_float_to_uint16(xnp)
self.inputs = {
'X': xnp,
'Index': np.array(self.index).astype(self.index_type),
Expand Down
21 changes: 17 additions & 4 deletions test/xpu/test_scatter_op_xpu.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@
get_xpu_op_support_types,
type_dict_str_to_numpy,
)
from op_test import convert_float_to_uint16
from op_test_xpu import XPUOpTest

import paddle
Expand Down Expand Up @@ -116,17 +117,29 @@ def setUp(self):
)

if not hasattr(self, 'init_ref_np'):
self.ref_np = np.ones((3, 50)).astype(self.dtype)
self.ref_np = np.ones((3, 50)).astype(
self.dtype if self.dtype != np.uint16 else np.float32
)
self.index_np = np.array([1, 2]).astype(self.index_type)
self.updates_np = np.random.random((2, 50)).astype(self.dtype)
self.updates_np = np.random.random((2, 50)).astype(
self.dtype if self.dtype != np.uint16 else np.float32
)
self.output_np = np.copy(self.ref_np)
self.output_np[self.index_np] = self.updates_np
else:
self.ref_np = self.init_ref_np.astype(self.dtype)
self.ref_np = self.init_ref_np.astype(
self.dtype if self.dtype != np.uint16 else np.float32
)
self.index_np = self.init_index_np.astype(self.index_type)
self.updates_np = self.init_updates_np.astype(self.dtype)
self.updates_np = self.init_updates_np.astype(
self.dtype if self.dtype != np.uint16 else np.float32
)
self.output_np = self.init_output_np.astype(self.dtype)

if self.dtype == np.uint16:
self.ref_np = convert_float_to_uint16(self.ref_np)
self.updates_np = convert_float_to_uint16(self.updates_np)

self.inputs = {
'X': self.ref_np,
'Ids': self.index_np,
Expand Down
Loading