Contents
・Installing tensorflow on a PC with RTX3060Ti
Introduction
この記事は機械学習：RTX3060TiとTensorflow2のGPU版の記事のなんちゃって英語版です。
私がインストールした時は、RTX3060Tiを積んだPCにtensorflowをインストールしたという情報が少なく苦労しました。どこかの誰かの役に立てばいいな、と一応つたない英語でも記録を残しておこうと思います。
I installed Tensorflow on a computer with an RTX3060Ti.
This article is a reminder, describing the installation process.
I installed tensorflow_gpu-2.3.0.
The version of tensorflow is 2, and it supports GPUs.
List of installed software
Anaconda (Python will also be installed when Anaconda is installed)
NVIDIA driver
CUDA
cuDNN
tensorflow_gpu
It is important to install the corresponding version of each software.
（It was a little difficult for me and I made a several times.）
Reference website:TensorFlow Official Website
(On the left side of the website, you can select Linux /macOS / Windows.)
The development environment I have built is as follows.
・Python：3.8.8
(After installing Anaconda, it was necessary to set the Python version when building the virtual environment.)
・NVIDIA driver：Latest version at the time of installation
・CUDA：10.1
・cuDNN：7.6
・tensorflow_gpu：2.3.0
What I struggled with…
Write a program that works with tensorflow2
I had previously created programs that worked with tensorflow1. Now that I have installed tensorflow2, I had to modify the programs a bit. It worked when I changed “tf.” to “tf.compat.v1”.
Reference website：TensorFlow Official Website（https://www.tensorflow.org)
Library version
It seems to be a good idea to have the latest versions of the libraries used in the program.
For all libraries that are not directly related to the installation of tensorflow2, it might be better to use the latest version. The other day, I tried to run a program I had created a long time ago, but the version of the library used in the program was out of date and I could not run it.
翻訳ソフトのおかげで英語力が絶望的な私でも英文ぽい文章を作成することができました。
Thanks to the translation software, I was able to create English sentences even though I am not good at English.
This is a blog post about accelerating computation speed using GPUs. I have successfully accelerated the execution speed of LSTM.
コメント