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.

external.py 1.6 kB

1234567891011121314151617181920212223242526272829303132333435363738394041424344
  1. # -*- coding: utf-8 -*-
  2. # MegEngine is Licensed under the Apache License, Version 2.0 (the "License")
  3. #
  4. # Copyright (c) 2014-2020 Megvii Inc. All rights reserved.
  5. #
  6. # Unless required by applicable law or agreed to in writing,
  7. # software distributed under the License is distributed on an
  8. # "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  9. # pylint: disable=too-many-lines
  10. from typing import List
  11. from ..tensor import Tensor
  12. def cambricon_subgraph(
  13. inputs: List[Tensor], data: bytes, symbol: str, tensor_dim_mutable: bool,
  14. ) -> List[Tensor]:
  15. """Loads a serialized Cambricon subgraph (i.e. cnrtModel_t) and
  16. execute the operations defined in the subgraph.
  17. :param inputs: list of input tensors of the subgraph.
  18. :param data: the serialized subgraph.
  19. :param symbol: the name of the function in the subgraph.
  20. The function is corresponding to a cnmlFusionOp
  21. which is added to the cnmlModel_t/cnrtModel_t.
  22. :param tensor_dim_mutable: whether the input tensors' shapes are mutalbe
  23. in cnrtModel_t.
  24. """
  25. raise NotImplementedError
  26. def extern_opr_subgraph(
  27. inputs, output_shapes: List[tuple], dump_name: str, dump_data: bytes,
  28. ) -> List[Tensor]:
  29. """Loads a serialized extern opr subgraph and fake execute the operator.
  30. :param inputs: tensor or list of input tensors.
  31. :param output_shapes: the output shapes.
  32. :param dump_name: the serialized subgraph name.
  33. :param dump_data: the serialized subgraph.
  34. :return: list of tensors.
  35. """
  36. raise NotImplementedError

MegEngine 安装包中集成了使用 GPU 运行代码所需的 CUDA 环境,不用区分 CPU 和 GPU 版。 如果想要运行 GPU 程序,请确保机器本身配有 GPU 硬件设备并安装好驱动。 如果你想体验在云端 GPU 算力平台进行深度学习开发的感觉,欢迎访问 MegStudio 平台