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_编译pytorch arm库

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

export LD_LIBRARY_PATH=/usr/local/lib64:/usr/local/lib:/usr/lib64:/usr/lib
/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 "<stdin>", line 1, in <module>
  File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/__init__.py", line 1465, in <module>
    from . import _meta_registrations
  File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_meta_registrations.py", line 7, in <module>
    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 <module>
    import torch._decomp.decompositions
  File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_decomp/decompositions.py", line 10, in <module>
    import torch._prims as prims
  File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_prims/__init__.py", line 33, in <module>
    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 <module>
    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 <module>
    from torch._guards import Source
  File "/usr/local/Python-3.10.12/lib/python3.10/site-packages/torch/_guards.py", line 78, in <module>
    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

[root@ceph3 pytorch-libs]# /usr/local/Python-3.10.12/bin/python3 -m pip install *.whl
Processing ./filelock-3.12.3-py3-none-any.whl
Processing ./Jinja2-3.1.2-py3-none-any.whl
Processing ./MarkupSafe-2.1.3-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
Processing ./mpmath-1.3.0-py3-none-any.whl
Processing ./networkx-3.1-py3-none-any.whl
Processing ./sympy-1.12-py3-none-any.whl
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)
mpmath is already installed with the same version as the provided wheel. Use --force-reinstall to force an installation of the wheel.
sympy is already installed with the same version as the provided wheel. Use --force-reinstall to force an installation of the wheel.
Installing collected packages: networkx, MarkupSafe, filelock, Jinja2
Successfully installed Jinja2-3.1.2 MarkupSafe-2.1.3 filelock-3.12.3 networkx-3.1
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
 
[notice] A new release of pip is available: 23.0.1 -> 23.2.1
[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