You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

conv_backprop_input.py 1.3 kB

5 years ago
12345678910111213141516171819202122232425262728293031323334
  1. #!/usr/bin/env python3
  2. # coding: utf-8
  3. # Copyright 2019 Huawei Technologies Co., Ltd
  4. #
  5. # Licensed under the Apache License, Version 2.0 (the "License");
  6. # you may not use this file except in compliance with the License.
  7. # You may obtain a copy of the License at
  8. #
  9. # http://www.apache.org/licenses/LICENSE-2.0
  10. #
  11. # Unless required by applicable law or agreed to in writing, software
  12. # distributed under the License is distributed on an "AS IS" BASIS,
  13. # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  14. # See the License for the specific language governing permissions and
  15. # limitations under the License.
  16. """conv_backprop_input"""
  17. from akg.ops.nn import conv_backprop_input
  18. def Conv2DBackpropInput(out_backprop, input_sizes, filter, filter_shape, pad_list, stride=1, dilation=1):
  19. """back propagation of 2d convolution on input"""
  20. if len(pad_list) != 4:
  21. raise IndexError("Length of pad must be equal 4")
  22. pad_ = pad_list
  23. data = []
  24. data.append(out_backprop)
  25. data.append(filter)
  26. fmap_shape = input_sizes
  27. filter_shape = filter_shape
  28. stride_ = [stride, stride]
  29. dilation_ = [dilation, dilation]
  30. return conv_backprop_input.conv_backprop_input(data, fmap_shape, filter_shape, pad_, stride_, dilation_)

AKG(Auto Kernel Generator)对深度神经网络中的算子进行优化,并提供特定模式下的算子自动融合功能。AKG与MindSpore的图算融合功能协同工作,可提升在不同硬件后端上运行网络的性能。