在WSL中通过Conda配置TensorFlow GPU加速遇阻求助

无Conda的WSL操作尝试 #

• 安装CUDA，执行nvcc -V显示版本为11.5.r11.5
• 创建并激活虚拟环境
• 安装TensorFlow
• 运行测试代码后检测不到GPU（设备实际配备GPU），测试代码如下：

import tensorflow as tf
from tensorflow.python.platform import build_info as build
# 检查TensorFlow能否访问GPU
physical_devices = tf.config.list_physical_devices('GPU')
print("Num GPUs Available: ", len(physical_devices))
if physical_devices:
    print("TensorFlow GPU details:")
    for gpu in physical_devices:
        print(gpu)
else:
    print("No GPUs detected by TensorFlow.")
print(tf.test.is_built_with_cuda())
print(build.build_info['cuda_version'])

基于Conda安装CUDA和cuDNN的WSL操作尝试 #

参考Anaconda Linux安装文档及cuDNN Conda安装指南，执行以下步骤：

• curl -O https://repo.anaconda.com/archive/Anaconda3-2024.02-1-Linux-x86_64.sh
• bash ~/Downloads/Anaconda3-2024.02-1-Linux-x86_64.sh
• 创建conda环境：conda create --name cudnn_env python=3.8
• 激活环境：conda activate cudnn_env
• 安装cuDNN：conda install -c conda-forge cudnn
• 运行测试代码python TfGpuTest.py，输出如下：

2024-05-27 17:23:19.552732: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: SSE4.1 SSE4.2 AVX AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
Num GPUs Available:  0
No GPUs detected by TensorFlow.
False
Traceback (most recent call last):
  File "TfGpuTest.py", line 18, in <module>
    print(build.build_info['cuda_version'])
KeyError: 'cuda_version'

补充尝试 #

原环境Python3.8对应的TensorFlow版本较旧，于是创建Python3.12的新环境，执行conda install tensorflow时出现版本兼容错误，无匹配版本（未找到TensorFlow 2.15）；降级至Python3.11.9后成功安装TensorFlow 2.12.0，但GPU检测问题仍未解决。

内容的提问来源于stack exchange，提问作者plutownium