尝鲜Ubuntu22.04 下 PaddlePaddle-GPU安装踩坑记-阿里云开发者社区

众所周知，之前飞桨一直不支持ubuntu22.04,只有比较老旧的ubuntu18.04, 16.04等老系统。

今天为了部署，特意安装ubuntu系统，装完ubuntu18.04才赫然发现，已经支持CUDA11.7了，突然想到这个版本是支持ubuntu22.04的，随即立刻下载新版本ubuntu，重新再次安装。

总结，今天安装了2次ubuntu。

ubuntu18.04 不自带显卡驱动，3060显卡下分辨率800*600很难搞。
ubuntu22.04 默认情况下不支持我的博通网卡（当年为了黑苹果特意买的）

0.硬件情况

3060显卡一枚
9400cpu一枚
分区固态500Gb
其他硬件

1.安装前系统准备

装完系统，网卡不支持，通过手机USB网络共享上网（android系统自带）
点击软件及更新，设置更新源为阿里源，估计更新速度会快
点击 “Additional Drivers”，选择nvidia-driver-520(proprietary)（英伟达显卡专用驱动）
此外，选择Broadcom Wireless Network Adapter(无线网卡驱动)

更新ing，持续4个多小时，我的小水管流量心疼啊。。。。。。

2.安装 miniconda

打开 tuna.moe 找到 miniconda ，下载并安装。
打开 tuna.moe/oh-my-tuna ，按指示操作换源（conda、pip）

3.安装 PaddlePaddle-GPU

打开 www.paddlepaddle.org.cn/ 官网，选择 conda、linux、2.4、cuda 11.7 。

3.1创建虚拟环境

首先根据具体的 Python 版本创建 Anaconda 虚拟环境，PaddlePaddle 的 Anaconda 安装支持 3.6 - 3.10 版本的 Python 安装环境。

conda create -n paddle_env python=YOUR_PY_VER

3.2 进入 Anaconda 虚拟环境

conda activate paddle_env

3.3 安装paddlepaddle-gpu=

对于 CUDA 11.7，需要搭配 cuDNN 8.4.1(多卡环境下 NCCL>=2.7)，安装命令为:

conda install paddlepaddle-gpu==2.4.0 cudatoolkit=11.7 -c https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/Paddle/

切记：

不要按照官网：

conda install paddlepaddle-gpu==2.4.0 cudatoolkit=11.7 -c https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/Paddle/ -c conda-forge

操作，要删掉 -c conda-forge，否则安装会特别慢，说不定还会中断，切记。。。。。。

4.排除bug操作

4.1 ImportError: libpython3.9.so.1.0: cannot open shared object file: No such file or directory

import paddle

>>> import paddle
Error: Can not import paddle core while this file exists: /home/livingbody/miniconda3/envs/p2/lib/python3.9/site-packages/paddle/fluid/libpaddle.so
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/livingbody/miniconda3/envs/p2/lib/python3.9/site-packages/paddle/__init__.py", line 25, in <module>
    from .framework import monkey_patch_variable
  File "/home/livingbody/miniconda3/envs/p2/lib/python3.9/site-packages/paddle/framework/__init__.py", line 17, in <module>
    from . import random  # noqa: F401
  File "/home/livingbody/miniconda3/envs/p2/lib/python3.9/site-packages/paddle/framework/random.py", line 16, in <module>
    import paddle.fluid as fluid
  File "/home/livingbody/miniconda3/envs/p2/lib/python3.9/site-packages/paddle/fluid/__init__.py", line 36, in <module>
    from . import framework
  File "/home/livingbody/miniconda3/envs/p2/lib/python3.9/site-packages/paddle/fluid/framework.py", line 37, in <module>
    from . import core
  File "/home/livingbody/miniconda3/envs/p2/lib/python3.9/site-packages/paddle/fluid/core.py", line 304, in <module>
    raise e
  File "/home/livingbody/miniconda3/envs/p2/lib/python3.9/site-packages/paddle/fluid/core.py", line 249, in <module>
    from . import libpaddle
ImportError: libpython3.9.so.1.0: cannot open shared object file: No such file or directory

处理办法：

如下所示，各位根据自己路径进行修改。

(p2) livingbody@gaint:~/miniconda3/envs/p2/lib$ sudo cp libpython3.9.so.1.0 /usr/lib
[sudo] livingbody 的密码： 
(p2) livingbody@gaint:~/miniconda3/envs/p2/lib$ sudo cp libpython3.9.so.1.0 /usr/lib64

4.2 PreconditionNotMetError: Cannot load cudnn shared library. Cannot invoke method cudnnGetVersion.

PreconditionNotMetError: Cannot load cudnn shared library. Cannot invoke method cudnnGetVersion.

>>> paddle.utils.run_check()
Running verify PaddlePaddle program ... 
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/utils/install_check.py", line 269, in run_check
    _run_static_single(use_cuda, use_xpu, use_npu)
  File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/utils/install_check.py", line 173, in _run_static_single
    exe.run(startup_prog)
  File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/executor.py", line 1463, in run
    six.reraise(*sys.exc_info())
  File "/home/livingbody/miniconda3/lib/python3.9/site-packages/six.py", line 703, in reraise
    raise value
  File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/executor.py", line 1450, in run
    res = self._run_impl(program=program,
  File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/executor.py", line 1661, in _run_impl
    return new_exe.run(scope, list(feed.keys()), fetch_list,
  File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/executor.py", line 631, in run
    tensors = self._new_exe.run(scope, feed_names,
RuntimeError: In user code:
    File "<stdin>", line 1, in <module>
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/utils/install_check.py", line 269, in run_check
      _run_static_single(use_cuda, use_xpu, use_npu)
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/utils/install_check.py", line 159, in _run_static_single
      input, out, weight = _simple_network()
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/utils/install_check.py", line 33, in _simple_network
      weight = paddle.create_parameter(
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/layers/tensor.py", line 152, in create_parameter
      return helper.create_parameter(attr, shape, convert_dtype(dtype), is_bias,
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/layer_helper_base.py", line 381, in create_parameter
      self.startup_program.global_block().create_parameter(
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/framework.py", line 3965, in create_parameter
      initializer(param, self)
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/initializer.py", line 56, in __call__
      return self.forward(param, block)
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/initializer.py", line 184, in forward
      op = block.append_op(type="fill_constant",
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/framework.py", line 4017, in append_op
      op = Operator(
    File "/home/livingbody/miniconda3/lib/python3.9/site-packages/paddle/fluid/framework.py", line 2858, in __init__
      for frame in traceback.extract_stack():
    PreconditionNotMetError: Cannot load cudnn shared library. Cannot invoke method cudnnGetVersion.
      [Hint: cudnn_d_handle should not be null.] (at /paddle/paddle/phi/backends/dynload/cudnn.cc:60)
      [operator < fill_constant > error]

解决办法：根据命令所知，需要的cuda、cudnn都已经安装，出现这个问题是找不到对应的动态库，所以要针对性处理。

4.2.1 mkdir

创建存放动态库的文件夹

mkdir /usr/local/cuda/lib64 -rf

4.2.2 拷贝 cuda 的 lib

拷贝动态库到lib

~/miniconda3/pkgs/cudatoolkit-11.7.0-hd8887f6_10/lib$ sudo cp * /usr/local/cuda/lib64 -rf

4.2.3 拷贝 cudnn 的 lib

覆盖性拷贝，同手动安装cudnn操作

~/miniconda3/pkgs/cudnn-8.4.1.50-hed8a83a_0/lib$ sudo cp * /usr/local/cuda/lib64/ -rf

4.2.4 设置 LD_LIBRARY_PATH 环境变量

编辑 .bahsrc

gedit ~/.bashrc

末尾添加

export LD_LIBRARY_PATH="/usr/local/cuda/lib64"

5.安装完毕

如果大家觉得有用，欢迎来个赞

尝鲜Ubuntu22.04 下 PaddlePaddle-GPU安装踩坑记