• pytorch-v2.0.1 cuda arm64 aarch64 torch 2.0.1+cu118 源码编译笔记【2】验证cuda安装 成功


    接上篇

    pytorch-v2.0.1 cuda arm64 aarch64 torch 2.0.1+cu118 源码编译笔记_hkNaruto的博客-CSDN博客

    由于采用/usr/local/bin/gcc编译,先设置LD_LIBRARY_PATH,再启动python3

    1. export LD_LIBRARY_PATH=/usr/local/lib64:/usr/local/lib:/usr/lib64:/usr/lib
    2. /usr/local/Python-3.10.12/bin/python3

    import torch报错

    # /usr/local/Python-3.10.12/bin/python3
    Python 3.10.12 (main, Sep  4 2023, 10:01:29) [GCC 9.2.0] on linux
    Type "help", "copyright", "credits" or "license" for more information.
    >>> import torch
    No sympy found
    Traceback (most recent call last):
      File "", line 1, in
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/__init__.py", line 1465, in
        from . import _meta_registrations
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_meta_registrations.py", line 7, in
        from torch._decomp import _add_op_to_registry, global_decomposition_table, meta_table
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_decomp/__init__.py", line 169, in
        import torch._decomp.decompositions
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_decomp/decompositions.py", line 10, in
        import torch._prims as prims
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_prims/__init__.py", line 33, in
        from torch._subclasses.fake_tensor import FakeTensor, FakeTensorMode
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_subclasses/__init__.py", line 3, in
        from torch._subclasses.fake_tensor import (
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_subclasses/fake_tensor.py", line 13, in
        from torch._guards import Source
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_guards.py", line 78, in
        class ShapeGuard(NamedTuple):
      File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_guards.py", line 79, in ShapeGuard
        expr: sympy.Expr
    NameError: name 'sympy' is not defined
     

    安装sympy

    在线

    /usr/local/Python-3.10.12/bin/python3 -m pip install sympy

    离线

    1. [root@ceph3 pytorch-libs]# /usr/local/Python-3.10.12/bin/python3 -m pip install *.whl
    2. Processing ./filelock-3.12.3-py3-none-any.whl
    3. Processing ./Jinja2-3.1.2-py3-none-any.whl
    4. Processing ./MarkupSafe-2.1.3-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
    5. Processing ./mpmath-1.3.0-py3-none-any.whl
    6. Processing ./networkx-3.1-py3-none-any.whl
    7. Processing ./sympy-1.12-py3-none-any.whl
    8. Requirement already satisfied: typing-extensions>=4.7.1 in /usr/local/Python-3.10.12/lib/python3.10/site-packages (from filelock==3.12.3) (4.7.1)
    9. mpmath is already installed with the same version as the provided wheel. Use --force-reinstall to force an installation of the wheel.
    10. sympy is already installed with the same version as the provided wheel. Use --force-reinstall to force an installation of the wheel.
    11. Installing collected packages: networkx, MarkupSafe, filelock, Jinja2
    12. Successfully installed Jinja2-3.1.2 MarkupSafe-2.1.3 filelock-3.12.3 networkx-3.1
    13. WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
    14. [notice] A new release of pip is available: 23.0.1 -> 23.2.1
    15. [notice] To update, run: /usr/local/Python-3.10.12/bin/python3 -m pip install --upgrade pip

    import torch成功,测试cuda是否可用

    参考:

    PyTorch快速安装并验证GPU是否可用_pytorch测试gpu_ai_patch的博客-CSDN博客

  • 相关阅读:
    【JAVA EE】详解单点登录
    使用ABP SignalR重构消息服务(二)
    【网络通信三】研华网关Modbus服务设置
    我们有一个新策划,请查收!
    linux检测系统是否被入侵(上)
    2023.11.6联赛总结
    devops学习Day2-单元测试jacoco
    集合的增删改查?--Python
    7.动态SQL
    【服务器数据恢复】HP StorageWorks系列服务器RAID5两块盘离线的数据恢复
  • 原文地址:https://blog.csdn.net/hknaruto/article/details/132692665