未验证 提交 6be3ee26 编写于 作者: Z zhangyikun02 提交者: GitHub

remove fp16 support of depthwise_conv2d and add unittest for depthwise_conv2d, test=kunlun (#43483)

上级 4002f320
......@@ -173,12 +173,8 @@ REGISTER_OP_XPU_KERNEL(
paddle::platform::float16>);
REGISTER_OP_XPU_KERNEL(
depthwise_conv2d,
ops::GemmConvXPUKernel<paddle::platform::XPUDeviceContext, float>,
ops::GemmConvXPUKernel<paddle::platform::XPUDeviceContext,
paddle::platform::float16>);
ops::GemmConvXPUKernel<paddle::platform::XPUDeviceContext, float>);
REGISTER_OP_XPU_KERNEL(
depthwise_conv2d_grad,
ops::GemmConvGradXPUKernel<paddle::platform::XPUDeviceContext, float>,
ops::GemmConvGradXPUKernel<paddle::platform::XPUDeviceContext,
paddle::platform::float16>);
ops::GemmConvGradXPUKernel<paddle::platform::XPUDeviceContext, float>);
#endif
......@@ -79,11 +79,9 @@ XPUOpMap& get_kl2_ops() {
{"conv2d_transpose",
XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
{"depthwise_conv2d_grad",
XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
pOpKernelType(vartype::FP16, XPUPlace())})},
XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
{"depthwise_conv2d",
XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
pOpKernelType(vartype::FP16, XPUPlace())})},
XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
{"dropout_grad",
XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
pOpKernelType(vartype::FP16, XPUPlace())})},
......
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import print_function
import sys
sys.path.append("..")
import unittest
import numpy as np
import paddle
paddle.enable_static()
import paddle.fluid.core as core
import paddle.fluid as fluid
from op_test_xpu import XPUOpTest
from paddle.fluid import Program, program_guard
from test_conv2d_op_xpu import XPUTestConv2DOp, XPUTestConv2DOp_v2
from xpu.get_test_cover_info import create_test_class, get_xpu_op_support_types, XPUOpTestWrapper
class XPUTestDepthwiseConv2DOp(XPUOpTestWrapper):
def __init__(self):
self.op_name = 'depthwise_conv2d'
self.use_dynamic_create_class = False
class TestDepthwiseConv(XPUTestConv2DOp.TestConv2DOp):
def init_test_case(self):
self.use_cuda = False
self.pad = [1, 1]
self.stride = [2, 2]
self.input_size = [2, 12, 5, 5] # NCHW
self.groups = 12
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [12, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
class TestDepthwiseConv2(XPUTestConv2DOp.TestConv2DOp):
def init_test_case(self):
self.use_cuda = False
self.pad = [1, 1]
self.stride = [1, 1]
self.input_size = [2, 12, 5, 5] # NCHW
self.groups = 12
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [12, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
class TestDepthwiseConv3(XPUTestConv2DOp.TestConv2DOp):
def init_test_case(self):
self.use_cuda = False
self.pad = [1, 1]
self.stride = [1, 1]
self.input_size = [2, 24, 5, 5] # NCHW
self.groups = 24
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [24, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
class TestDepthwiseConvWithDilation(XPUTestConv2DOp.TestConv2DOp):
def init_test_case(self):
self.use_cuda = False
self.pad = [1, 1]
self.stride = [2, 2]
self.input_size = [2, 24, 5, 5] # NCHW
self.groups = 24
self.dilations = [2, 2]
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [24, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
class TestDepthwiseConvWithDilation2(XPUTestConv2DOp.TestConv2DOp):
def init_test_case(self):
self.use_cuda = False
self.pad = [1, 1]
self.stride = [1, 1]
self.input_size = [2, 24, 5, 5] # NCHW
self.groups = 24
self.dilations = [2, 2]
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [24, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
class XPUTestDepthwiseConv2DOp_v2(XPUOpTestWrapper):
def __init__(self):
self.op_name = 'depthwise_conv2d'
self.use_dynamic_create_class = False
class TestDepthwiseConv_AsyPadding(XPUTestConv2DOp_v2.TestConv2DOp_v2):
def init_test_case(self):
self.use_cuda = False
self.stride = [2, 2]
self.input_size = [2, 12, 5, 5] # NCHW
self.groups = 12
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [12, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
def init_paddings(self):
self.pad = [1, 1, 0, 1]
self.padding_algorithm = "EXPLICIT"
class TestDepthwiseConv2_AsyPadding(XPUTestConv2DOp_v2.TestConv2DOp_v2):
def init_test_case(self):
self.use_cuda = False
self.stride = [1, 1]
self.input_size = [2, 12, 5, 5] # NCHW
self.groups = 12
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [12, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
def init_paddings(self):
self.pad = [0, 1, 0, 2]
self.padding_algorithm = "EXPLICIT"
class TestDepthwiseConv3_AsyPadding(XPUTestConv2DOp_v2.TestConv2DOp_v2):
def init_test_case(self):
self.use_cuda = False
self.stride = [1, 1]
self.input_size = [2, 24, 5, 5] # NCHW
self.groups = 24
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [24, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
def init_paddings(self):
self.pad = [1, 1, 0, 0]
self.padding_algorithm = "EXPLICIT"
class TestDepthwiseConvWithDilation_AsyPadding(
XPUTestConv2DOp_v2.TestConv2DOp_v2):
def init_test_case(self):
self.use_cuda = False
self.pad = [1, 1]
self.stride = [2, 2]
self.input_size = [2, 24, 5, 5] # NCHW
self.groups = 24
self.dilations = [2, 2]
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [24, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
def init_paddings(self):
self.pad = [1, 1, 2, 1]
self.padding_algorithm = "EXPLICIT"
class TestDepthwiseConvWithDilation2_AsyPadding(
XPUTestConv2DOp_v2.TestConv2DOp_v2):
def init_test_case(self):
self.use_cuda = True
self.pad = [1, 1]
self.stride = [1, 1]
self.input_size = [2, 24, 5, 5] # NCHW
self.groups = 24
self.dilations = [2, 2]
assert np.mod(self.input_size[1], self.groups) == 0
f_c = self.input_size[1] // self.groups
self.filter_size = [24, f_c, 3, 3]
self.op_type = "depthwise_conv2d"
def init_paddings(self):
self.pad = [0, 1, 1, 0]
self.padding_algorithm = "EXPLICIT"
support_types = get_xpu_op_support_types('depthwise_conv2d')
for stype in support_types:
create_test_class(globals(), XPUTestDepthwiseConv2DOp, stype)
create_test_class(globals(), XPUTestDepthwiseConv2DOp_v2, stype)
#### depthwise conv2d
# create_test_padding_SAME_class(TestDepthwiseConv_AsyPadding)
# create_test_padding_SAME_class(TestDepthwiseConvWithDilation_AsyPadding)
# create_test_padding_SAME_class(TestDepthwiseConvandFuse_AsyPadding)
# create_test_padding_SAME_class(TestDepthwiseConvWithDilationandFuse_AsyPadding)
# create_test_padding_VALID_class(TestDepthwiseConv_AsyPadding)
# create_test_padding_VALID_class(TestDepthwiseConvWithDilation_AsyPadding)
# create_test_padding_VALID_class(TestDepthwiseConvandFuse_AsyPadding)
# create_test_padding_VALID_class(TestDepthwiseConvWithDilationandFuse_AsyPadding)
#### channel last
# create_test_channel_last_class(TestDepthwiseConv_AsyPadding)
# create_test_channel_last_class(TestDepthwiseConvWithDilation2_AsyPadding)
# create_test_channel_last_class(TestDepthwiseConvandFuse_AsyPadding)
# create_test_channel_last_class(TestDepthwiseConvWithDilationandFuse_AsyPadding)
if __name__ == '__main__':
unittest.main()
Markdown is supported
0% .
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