如何在 Windows 上编译具有 GPU 支持的 Tensorflow Java API？
[英]How to Compile Tensorflow Java API with GPU Support on Windows?
问题描述
I have some models build with Tensorflow Java API running on Windows.
我有一些使用在 Windows 上运行的 Tensorflow Java API 构建的模型。 But, I can't run these models on GPU.但是，我无法在 GPU 上运行这些模型。
I am trying to compile Tensorflow Java API on Windows following the steps described here .我正在尝试按照此处描述的步骤在 Windows 上编译 Tensorflow Java API。 But it does not have GPU support or any complementar instruction for that.但它没有 GPU 支持或任何补充说明。
Please, could help to solve this issues?请问能帮忙解决这个问题吗？
2 个解决方案
解决方案1
2 已采纳 2020-02-05 14:08:34
Hello I do not know if the issue is still relevant, but Tensorflow already provides GPU support for Windows.您好，我不知道这个问题是否仍然相关，但 Tensorflow 已经为 Windows 提供了 GPU 支持。 Just make sure you have the correct version of CUDA installed for your version of Java TF and that your GPU has the minimum computing capability.只需确保为您的 Java TF 版本安装了正确版本的 CUDA，并且您的 GPU 具有最低计算能力。
The CUDA versions might be a little tricky. CUDA 版本可能有点棘手。 I tried with both tf 1.12 and tf 1.15.我尝试了 tf 1.12 和 tf 1.15。 Tf 1.12 worked with CUDA 9.0 and tf 1.15 worked with CUDA 10.0 tf 1.12 适用于 CUDA 9.0，tf 1.15 适用于 CUDA 10.0
解决方案2
0 2017-08-23 23:39:49
Windows+Java+GPU is not supported yet.尚不支持 Windows+Java+GPU。 Here is a related github issue https://github.com/tensorflow/tensorflow/issues/10708这是一个相关的 github 问题https://github.com/tensorflow/tensorflow/issues/10708
声明:本站的技术帖子网页，遵循CC BY-SA 4.0协议，如果您需要转载，请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.