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.

pack_kernel.h 1.8 kB

5 years ago
5 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445
  1. /**
  2. * Copyright 2020 Huawei Technologies Co., Ltd
  3. *
  4. * Licensed under the Apache License, Version 2.0 (the "License");
  5. * you may not use this file except in compliance with the License.
  6. * You may obtain a copy of the License at
  7. *
  8. * http://www.apache.org/licenses/LICENSE-2.0
  9. *
  10. * Unless required by applicable law or agreed to in writing, software
  11. * distributed under the License is distributed on an "AS IS" BASIS,
  12. * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13. * See the License for the specific language governing permissions and
  14. * limitations under the License.
  15. */
  16. #ifndef GE_GRAPH_PASSES_FOLDING_KERNEL_PACK_KERNEL_H_
  17. #define GE_GRAPH_PASSES_FOLDING_KERNEL_PACK_KERNEL_H_
  18. #include <vector>
  19. #include "inc/kernel.h"
  20. namespace ge {
  21. /**
  22. * @ingroup ge
  23. * @brief Pack operator processing
  24. * @author
  25. */
  26. class PackKernel : public Kernel {
  27. public:
  28. Status Compute(const ge::OpDescPtr op_desc_ptr, const std::vector<ge::ConstGeTensorPtr> &input,
  29. std::vector<ge::GeTensorPtr> &v_output) override;
  30. private:
  31. Status ValidateKernelParams(const ge::OpDescPtr &op_desc_ptr, const std::vector<ge::ConstGeTensorPtr> &input);
  32. Status ValidateInputs(const ge::OpDescPtr &op_desc_ptr, const std::vector<ge::ConstGeTensorPtr> &input);
  33. void ExpandDims(const int64_t axis, const std::vector<ge::ConstGeTensorPtr> &input, GeShape &final_shape);
  34. Status CopyOutputData(const GeShape &final_shape, const std::vector<ge::ConstGeTensorPtr> &input,
  35. ge::GeTensorPtr &output_ptr);
  36. int64_t n_ = 0; // count of inputs
  37. int64_t axis_ = 0; // axis to stack along.
  38. DataType data_type_; // data type of inputs
  39. };
  40. } // namespace ge
  41. #endif // GE_GRAPH_PASSES_FOLDING_KERNEL_PACK_KERNEL_H_

图引擎模块(GE)是MindSpore的一个子模块,其代码由C++实现,位于前端模块ME和底层硬件之间,起到承接作用。图引擎模块以ME下发的图作为输入,然后进行一系列的深度图优化操作,最后输出一张可以在底层硬件上高效运行的图。GE针对昇腾AI处理器的硬件结构特点,做了特定的优化工作,以此来充分发挥出昇腾AI处理器的强大算力。在进行模型训练/推理时,GE会被自动调用而用户并不感知。GE主要由GE API和GE Core两部分组成,详细的架构图如下所示