2020, TensorFlow 2.2 NVIDIA GPU (CUDA)/CPU, Keras, & Python 3.7 in Linux Ubuntu
HTML-код
- Опубликовано: 10 сен 2024
- In this video I show how to install Keras/Python/Tensorflow with an NVIDIA GPU on an Ubuntu system, assuming Linux as the desktop GUI. I install on a Lenovo ThinkPad P53.
I use the instructions here:
github.com/jef...
Lenovo P-series:
www.lenovo.com...
NVIDIA Quadro RTX 5000:
www.nvidia.com...
** Follow Me on Social Media!
GitHub: github.com/jef...
Twitter: / jeffheaton
Instagram: / jeffheatondotcom
Discord: / discord
Patreon: / jeffheaton
I learned neural nets from you 12-13 years ago with C# from scratch! There was no Tensorflow or data science craze using Python! Thank you so much. I had both of your books Using Neural Nets with C# lol! I remember those days so well. Neural Nets have come a long way since then...Thank you very much Jeff!
Wow....just wow! I have been trying to install GPU version of TF in my Linux OS for the past 3 days without any success. This video just made my day. Thank you very much!
Glad it helped!
Thank you! Out of many tutorials on RUclips, this one worked perfectly
I love how you simplified so many concepts I have learnt a lot listening to you.
If I may suggest something, please give us a video on real time prediction or deployment of a model.
Deployment is very much on my list. I will think about real time as well.
Thanks a lot, I have headache for installing tensorflow with cuda and cudnn since last year. Thanks a lot @Jeff Heaton
Nice tutorial sir, I wish I had seen this 2days back I was missing some steps but I am back on doctor now.
Thank you so much, an useful video
I am facing issue with opening jupyter notebook. It is throwing 500 Internal server error. Could you please help with it.
Thank you so much. I have tried it and everything is working good on my system
Very helpful sir thank you very much! Fantastic! I just installed Ubuntu 22.04 with Python 3.9. Everything still works, just need to update some version numbers. Thanks again!
Great Video Dr. Heaton.
Thank you sooo, much. This video was soooooooo helpful.
❤❤❤❤
So as long as we're in Linux system, we don't have to install cuda and cudnn by ourselves. Conda will do the job for us, am I right?
Correct! Actually on Windows, CONDA will do the same.... ruclips.net/video/PnK1jO2kXOQ/видео.html&ab_channel=JeffHeaton
Hi Jeff, thanks a lot for posting such an informative video. I just have a quick question: does this tutorial/setup assume the proprietary NVIDIA drivers have been install on the host OS?
Heads up for people trying this in 2021. If you follow the video exactly, it should work, I think the conda package for python 3.8 is broken but 3.7 is good
Is this for Ubuntu version 18 or does it work with version 20 aswell?
@@PontusThePilot works for 20 as well. I'm currently on 20
@@tmnic6971 okay thank you. Then I’ll try this tutorial.
I had a problem with another tutorial which used cuda 10 and I wasn’t able to install it, the computer automatically installed toolkit 11 even tho I downloaded the installer for version 10
@Kaustav Bakshi As Tm Nic said you have to use python 3.7. It works for me then and says “gpu avalanche”, I’ve tried my exact same setup but with 3.8 and it then said “not available”. My problem now is that tensorflow doesn’t support cuda 11.2 but that is another issue
@Kaustav Bakshi yep Pontus is correct. Although tensorflow supports 3.8, I think the conda package for 3.8 is broken but the 3.7 should work. Just specify 3.7 for now if you want the easy installation through conda
Thank you so much professor!
You’re the best Jeff
I have followed your video to install TensorFlow 2.2. When I list the packages in my newly created tensorflow virtual environment, it shows tensorflow version 2.2.0. However, when I try to import it in juypter notebook, it shows version 1.14. I am confused, it's not clear what I am missing here. Any suggestions would be great!
For those with the issue: "failed call to cuInit: CUDA_ERROR_UNKNOWN: unknown error", here is the fix.
Check this post: github.com/tensorflow/tensorflow/issues/32623 and follow the instructions:
cd /dev
sudo nano script.sh (copy/write contents of "Device Node Verification Script")
sudo chmod 777 script.sh
sudo ./script.sh
This one works for me.
There is a way to get working old gtx 650 ti as tensorflow gpu? I tried on windows 10 and Ubuntu but no success. Is this a waste of time to try this presently?
Hello sir, when I want to installing the miniconda in the specify location. What should command I use ?
You're awesome! Thanks.
Hi, thank you for this amazing tutorial; however, after using conda for tensorflow-gpu, I still get report that GPU is not available. How can I fix it ?
Buy a machine with a GPU
Muchas gracias por su ayuda. Saludos
de nada! 😃
I used this installation Guide on Ubuntu 18.04
Tensorflow does recognize my GPU but once i run a model i get the following error
UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
[[node model/stem_conv1/Conv2D (defined at :11) ]] [Op:__inference_train_function_45957]
Function call stack:
train_function
can anyone help me ?
Which GPU are you using?
And one more question:
will this same work in manjaro or other linux systems as well?
GOAT
If you have the time, I'd like to see a PyTorch version....please, as I'm exploring the fast.ai library. Thanks a bunch.
Sure, on my list.
My concern with a pytorch install video is that then we will get fewer videos from the Nvidia conference... keep up the good work Jeff!
yes, please, i really need help on this...
hi Jeff .. Would you be able to update this video please.. with tensorflow gpu 2.9, python 3.9 etc ?
Thank you for simplified information. Its really helpful. I followed the same steps and did installation but its showing PermissionError after installation like: [Errno 13] Permission denied: '/usr/share/jupyter/nbconvert/templates/conf.json', when I am opening jupyter notebook new . ipynb file. please reply how to overcome this error.
Is there a way to make conda not display (base) on every terminal?
use this:
$ conda config --set auto_activate_base false
and than use:
$ conda activate
to get into (base)
Thanks, that is very cool
Thanks for the instructions. I followed till end what was instructed. But after running code to verify GPU, I got output as : GPU is NOT AVAILABLE.
Please suggest. I have GPU : Nvidia Quadro P1000.
I have not checked it but is the Nvidia Quadro P 1000 compatible with Cuda?
I really need a tutorial like this for pytorch.... im going mad trying to make pytorch comunicate with the gpu, even though the drivers are installed and the gpu works just fine in everything else...
Working on that now, should be ready to post on Wed.
@@HeatonResearch i can`t thank you enough!
the tml file step is broken, when i enter the "conda env update --file tools.yml" it says
Warning: you have pip-installed dependencies in your environment file, but you do not list pip itself as one of your conda dependencies. Conda may not use the correct pip to install your packages, and they may end up in the wrong place. Please add an explicit pip dependency. I'm adding one for you, but still nagging you.
and this goes on to infinity and no install starts :(
@Mark Wetzel will try.
@Mark Wetzel worked great i added a -pip before -pip: worked like a charm
Have you ever done an nlp focused DL network with gpu ?
Not extensively, but might be something to look at in a future video.
Thank you for the tuto but unfortunately i have an error while importing tensorflow in Python:
(tensorflow) mct@mct-G3-3579:~$ python
Python 3.7.9 (default, Aug 31 2020, 12:42:55)
[GCC 7.3.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
>>> tf.config.list_physical_devices("GPU")
2020-11-04 18:34:39.222356: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
2020-11-04 18:34:41.047688: E tensorflow/stream_executor/cuda/cuda_driver.cc:313] failed call to cuInit: CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected
2020-11-04 18:34:41.047802: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:156] kernel driver does not appear to be running on this host (mct-G3-3579): /proc/driver/nvidia/version does not exist
[]
Can you please help me ? i am trying now for weekd to download Tensorflow correctly
How to update cuda