python - Tensorflow 2.3.0 CUDA Toolkit 10.1版不使用GPU
问题描述
我的 RTX2070 gpu 有 tensorflow 2.0 workig。我做了一个 Windows 更新,所以我可以使用 tf-nightly。不喜欢它所以卸载它并重新安装 tensorflow 2.3.0。运行以前在 GPU 上运行良好但没有使用 GPU 的以前的 python 代码。尝试了很多东西。终于重新开始了。重新安装 Anaconda,创建新环境。卸载 Cuda 工具包 10.1 并重新安装。在目录 c:\Tools 中安装了 cuDnn SDK 7.6。检查要包含的路径环境变量
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin;
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\extras\CUPTI\lib64;
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include;
C:\tools\cuda\bin;%PATH%
#then ran this code:
import tensorflow as tf
from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())
print(tf.__version__)
print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU')))
tf.test.is_gpu_available()
#I get the result
[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 15177607927005893519
, name: "/device:XLA_CPU:0"
device_type: "XLA_CPU"
memory_limit: 17179869184
locality {
}
incarnation: 4640072765546557805
physical_device_desc: "device: XLA_CPU device"
, name: "/device:XLA_GPU:0"
device_type: "XLA_GPU"
memory_limit: 17179869184
locality {
}
incarnation: 16675502319763286567
physical_device_desc: "device: XLA_GPU device"
]
2.3.0
Num GPUs Available: 0
False
tensorflow still does not use GPU. What an I missing?
also same problem using python 3.7.0 and same problem using tensorflow 2.0.0
解决方案
我发现如果在我的工作环境中使用 conda 我运行 conda install cudnn==7.6.4 可以让 tensorflow 识别 GPU,它与 CUDA 10.1.0 一起使用,anaconda 提示符中的结果消息是:
Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Collecting package metadata (repodata.json): done
Solving environment: done
## Package Plan ##
environment location: C:\Users\tfuser\anaconda3\envs\tf
added / updated specs:
- cudnn==7.6.4
The following packages will be downloaded:
package | build
---------------------------|-----------------
cudnn-7.6.4 | cuda10.1_0 179.3 MB
------------------------------------------------------------
Total: 179.3 MB
The following NEW packages will be INSTALLED:
cudatoolkit pkgs/main/win-64::cudatoolkit-10.1.243-h74a9793_0
cudnn pkgs/main/win-64::cudnn-7.6.4-cuda10.1_0
Proceed ([y]/n)? y
The following packages will be downloaded:
cudnn-7.6.4 | cuda10.1_0 179.3 MB
The following NEW packages will be INSTALLED:
cudatoolkit pkgs/main/win-64::cudatoolkit-10.1.243-h74a9793_0
cudnn pkgs/main/win-64::cudnn-7.6.4-cuda10.1_0
Proceed ([y]/n)? y
Downloading and Extracting Packages
cudnn-7.6.4 | 179.3 MB |
Preparing transaction: doneVerifying transaction: done
Executing transaction: done
推荐阅读
- r - R中的lapply函数:删除行上具有重复名称的列
- gml - 有什么方法可以使用 HSV 选择表面的颜色吗?
- r - 从分析中省略某些列
- php - SQL: select * from users: SQLSTATE[HY000] [1045] Access denied for user 'root'@'localhost' (使用密码: YES)
- angularjs - 如何使后退按钮在角度状态下工作
- twitter-bootstrap-3 - 透明模态标题引导程序 3
- sql - PostgreSQL 使用来自 2 个表的数据,没有一个共同的值,而是一个布尔值
- discord.js - 分配或删除角色时如何在 Discord 上编辑嵌入消息
- python - 为什么代码无法用下划线替换空格
- node.js - 为什么对象没有在 express.Router() 的路由处理程序的回调中传递,从不同的模块导出?