I hope this video helped you with your TensorFlow installation - it still surprises me how [unnecessarily] difficult it can be to get installed properly, especially with the GPU. Please reply and let me know which method worked for you!
I'm so glad that I could help!! Yes, TensorFlow is one of the most important packages you need to get started, and it's one of the hardest to get working properly... which method did you end up using from the video?
Hi Dustin first thank you for a logical step by step video listing different methods in one video. Question: did you add any paths to the Enviroment Variables in Windows ? thanks again 👌
You are golden! Thank you verry much for all the teachings it was a great help in my academic works. May you have many nice days - as many as you saved my days from tears and sorrows with this tutorial! Cheers!
Thank you very much for your guide! It helped me a lot to set up the latest tensorflow gpu version in windows. Finally my GPU is showing up when importing tensorflow.👍 However, I'm getting 3 warnings, saying "Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered" and 2 mores similar messages with cuDNN and cuBLAS. Have you ever encountered such messages?
Wow, this is great, I have a question though, what about working with notebooks? Is there a way to be able to run it from vscode? The kernel isn't explicitly shown prolly 'cuz it's in a container. I've seen some tutorials about using the "WSL" and "Dev Containers" extensions but am still experimenting as to how to make them all work together. Still, I'm thankful that I found a tutorial that could utilize newer versions of TF. Edit: This is with regards to the Docker TF Installation
Hi @daedalusrasmus3278 - glad that you found it helpful! Yes, there's a docker version of TensorFlow that uses notebooks as well, but to be honest I haven't tried using VScode with it through docker. To open a notebook version in a docker container, use this: docker run --gpus all -p 8888:8888 -it tensorflow/tensorflow:latest-gpu-jupyter ... and then you can access the notebook from a web browser.
I am extremely new to this space. I just started neural networks and my processing times in google colab has been humungous. I figured its because its not using my gpu eventhough i use the t4gpu runtime. So i went with this tutorial and downloaded the docker and completed every step and it works. But how can I now use my jupyter notebook file?
There’s a different docker file you can use that includes Jupyter notebook… scroll through the comments a bit and you’ll find the one where I responded to someone else… use that command
after installing tensorflow gpu how to access it from the jupyter notebook as there is jupyter pre install in my pc through ananconda and in same command line i installed tensorflow but i was anable to get it on my jupyter notebook how to do it please help me out
@@RELUvance After installing the tensorflow and using jupyter notebook, when I run the following code cell, I get the error - The kernel appears to have died. It will restart automatically. Do you have any solution for this? Code: linear_layer = tf.keras.layers.Dense(units=1, activation = 'linear', ) a1 = linear_layer(X_train[0].reshape(1,1)) print(a1)
As I understand there is no more version 2.10 available, as I get this (tf-gpu) PS C:\Users\edgar> pip install tensorflow==2.10 Defaulting to user installation because normal site-packages is not writeable ERROR: Could not find a version that satisfies the requirement tensorflow==2.10 (from versions: 2.16.0rc0, 2.16.1) ERROR: No matching distribution found for tensorflow==2.10
If you are installing them via conda or docker like I showed in the video you can install them both if you want because they will be in separate environments. I would only install the gpu version though if you are planning on running really compute intensive models. Otherwise the cpu version will be fine on its own.
i just wanted to know whether we have to use docker run - -gpus all -it -v $PWD:/working -w /working tensorflow/tensorflow:latest-gpu bash, every time we reopen the linux shell after closing it or ??is there another way to go about it ,also to run the req python script we have use the above mentioned cmd and then call the script ?or is there some other way to do this
There are actually many ways to work with docker containers but I tried to keep the video simple and just showed this one example. For example, you can actually instantiate a docker container as a "daemon" with this command: docker run --gpus all -dit --name tf -v $PWD:/working -w /working tensorflow/tensorflow:latest-gpu bash This will cause the container to be persistent even after stopping the process or restarting the computer, and can be re-started by using: docker attach tf (replacing tf with whatever name you used in the preceding command. To run a script immediately instead of launching the container as a shell you can replace "bash" as follows: docker run - -gpus all -it -v $PWD:/working -w /working tensorflow/tensorflow:latest-gpu python [script name here] Hope this helps! As you can see it gets complicated depending on what you want to do, and this was beyond what I wanted to cover in the video. Let me know if you think a separate video on using docker would be helpful, and I might be able to make one.
The 2nd method I shows in the video actually has python installed natively in Windows so you should be able to fire up VS and use it directly within windows. For the docker use-case, it may be a bit more challenging because python and tensorflow are inside of a docker container. It's probably still possible, but I haven't explored this myself.
When i tried to run the sample script to verify whether the docker works, it said: "docker: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?. See 'docker run --help'.", will this cause any issues?, i can always run it via 'sudo service docker start' tho
If you can run with sudo but not without, it sounds like setting up the user profile like I showed in the video didn’t work properly. Did you get any error messages when doing that? Try rebooting the computer too if you haven’t already.
I hope this video helped you with your TensorFlow installation - it still surprises me how [unnecessarily] difficult it can be to get installed properly, especially with the GPU. Please reply and let me know which method worked for you!
Liked, commented, and commenting here to tell you thank you from the bottom of my heart!
would ykou tell me its help anaconda shell
@@mdabubakarchowdhurysunny2846 sorry, I don’t understand what you are asking
@@RELUvance i want to ask you tensorflow gpu doesn"t install conda environment , would yoiu like to tell us how i used install conda
@mdabubakarchowdhurysunny2846 At 5:21 in the video it explains how to install miniconda, which is a “lite” version of the full anaconda.
You are a wizard, Man! That's the simple and full installation guides, that are so necessary for beginners, like me!👍👍👍
Explained using Multiple methods, Is the best part for me.
Thank you so much, I was struggling for 2 days before I found your tutorial.
Glad it helped!
You just helped me overcome the current, biggest roadblock I had to starting my ML/AI journey. I and my daughters thank you, endlessly.
I'm so glad that I could help!! Yes, TensorFlow is one of the most important packages you need to get started, and it's one of the hardest to get working properly... which method did you end up using from the video?
Hi Dustin first thank you for a logical step by step video listing different methods in one video. Question: did you add any paths to the Enviroment Variables in Windows ? thanks again 👌
Hi Peter, no, I didn’t make any changes to windows paths.
@@RELUvance Okay thank you !
You are golden! Thank you verry much for all the teachings it was a great help in my academic works. May you have many nice days - as many as you saved my days from tears and sorrows with this tutorial! Cheers!
I’m glad you enjoyed!
thats sad caz many new lib req new version of tensorflow or keras , btw it worked . tq sir!
Glad it worked for you!
Thank you so much, you saved me HAHAHAH
You're welcome!!
excellent...this video helped me a lot
Fantastic!
@@RELUvance sir, can you please show how to install Tensorflow+GPU on WSL2 environment without dockers help
Yes, I am planning to make a separate video on that.
Thank you very much for your guide! It helped me a lot to set up the latest tensorflow gpu version in windows. Finally my GPU is showing up when importing tensorflow.👍
However, I'm getting 3 warnings, saying "Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered" and 2 mores similar messages with cuDNN and cuBLAS. Have you ever encountered such messages?
Great video! Can I install TF in a venv on cmd prompt instead of conda?
Yes you can use venv too.
Thanks alot.. would have preferred if the video was a little bit clearer.. although it was very helpful and precise..
Thank you! That's the only tutorial that worked for me. However, I just wanna know if it is possible to run VScode in the method3/docker.
Thank you for this video...
No problem! Hope that it helped you.
Wow, this is great, I have a question though, what about working with notebooks? Is there a way to be able to run it from vscode? The kernel isn't explicitly shown prolly 'cuz it's in a container. I've seen some tutorials about using the "WSL" and "Dev Containers" extensions but am still experimenting as to how to make them all work together. Still, I'm thankful that I found a tutorial that could utilize newer versions of TF. Edit: This is with regards to the Docker TF Installation
Hi @daedalusrasmus3278 - glad that you found it helpful! Yes, there's a docker version of TensorFlow that uses notebooks as well, but to be honest I haven't tried using VScode with it through docker. To open a notebook version in a docker container, use this:
docker run --gpus all -p 8888:8888 -it tensorflow/tensorflow:latest-gpu-jupyter
... and then you can access the notebook from a web browser.
I am extremely new to this space. I just started neural networks and my processing times in google colab has been humungous. I figured its because its not using my gpu eventhough i use the t4gpu runtime. So i went with this tutorial and downloaded the docker and completed every step and it works. But how can I now use my jupyter notebook file?
There’s a different docker file you can use that includes Jupyter notebook… scroll through the comments a bit and you’ll find the one where I responded to someone else… use that command
finally a latest video . thank u dude
Yeah that’s why I made it… all the other videos are so out of date.
thank you
You're welcome!
I followed all these steps for CPU and GPU and I am still getting errors when I try to import tensorflow :(
it works well
i have a laptop with intel iris xe graphics, integrated gpu can i run tensorflow?
I have got all the installation done. how can i use it from jupyter notebook?
you are amazing
Which is the best method among these 3?
Option 2 is the best if you don’t need the absolute latest gpu version (most people).
after installing tensorflow gpu how to access it from the jupyter notebook as there is jupyter pre install in my pc through ananconda and in same command line i installed tensorflow but i was anable to get it on my jupyter notebook how to do it please help me out
I truly appreciate it.
I’m glad it helped!
After doing this, can i use tensorflow in jupyter lab?
Yes, if you run Jupyter from within the Tensorflow environment you created.
Why did you use python version 3.9 for the windows for tensorflow installation?
That’s the version that is officially supported by the version of tensorflow being installed.
@@RELUvance After installing the tensorflow and using jupyter notebook, when I run the following code cell, I get the error - The kernel appears to have died. It will restart automatically. Do you have any solution for this?
Code:
linear_layer = tf.keras.layers.Dense(units=1, activation = 'linear', )
a1 = linear_layer(X_train[0].reshape(1,1))
print(a1)
As I understand there is no more version 2.10 available, as I get this (tf-gpu) PS C:\Users\edgar> pip install tensorflow==2.10
Defaulting to user installation because normal site-packages is not writeable
ERROR: Could not find a version that satisfies the requirement tensorflow==2.10 (from versions: 2.16.0rc0, 2.16.1)
ERROR: No matching distribution found for tensorflow==2.10
Hi Edgar, are you using the command I used in the video? (pip install "tensorflow
👑
should I install TensorFlow native CPU and GPU versions together or only one of them is allowed?
If you are installing them via conda or docker like I showed in the video you can install them both if you want because they will be in separate environments. I would only install the gpu version though if you are planning on running really compute intensive models. Otherwise the cpu version will be fine on its own.
i just wanted to know whether we have to use docker run - -gpus all -it -v $PWD:/working -w /working tensorflow/tensorflow:latest-gpu bash, every time we reopen the linux shell after closing it or ??is there another way to go about it ,also to run the req python script we have use the above mentioned cmd and then call the script ?or is there some other way to do this
There are actually many ways to work with docker containers but I tried to keep the video simple and just showed this one example. For example, you can actually instantiate a docker container as a "daemon" with this command: docker run --gpus all -dit --name tf -v $PWD:/working -w /working tensorflow/tensorflow:latest-gpu bash
This will cause the container to be persistent even after stopping the process or restarting the computer, and can be re-started by using: docker attach tf (replacing tf with whatever name you used in the preceding command.
To run a script immediately instead of launching the container as a shell you can replace "bash" as follows: docker run - -gpus all -it -v $PWD:/working -w /working tensorflow/tensorflow:latest-gpu python [script name here]
Hope this helps! As you can see it gets complicated depending on what you want to do, and this was beyond what I wanted to cover in the video. Let me know if you think a separate video on using docker would be helpful, and I might be able to make one.
@@RELUvance thanks for the clarification
Also a video on using docker for python scripts via the vscode or any other ide would be helpful
i have a laptop with intel iris xe graphics, integrated gpu
You need an Nvidia graphics card for the GPU-enabled version to work.
how can I use this in visual studio community?
The 2nd method I shows in the video actually has python installed natively in Windows so you should be able to fire up VS and use it directly within windows. For the docker use-case, it may be a bit more challenging because python and tensorflow are inside of a docker container. It's probably still possible, but I haven't explored this myself.
its cpu version
In the video I explain how to install both the gpu and cpu versions. Maybe I don’t understand your comment?
@@RELUvance can you solve problem in my next comment
Install tensorflow Bible thank you!🎉 Method 3 help me a lot
So glad that it worked for you!
When i tried to run the sample script to verify whether the docker works, it said: "docker: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?.
See 'docker run --help'.", will this cause any issues?, i can always run it via 'sudo service docker start' tho
If you can run with sudo but not without, it sounds like setting up the user profile like I showed in the video didn’t work properly. Did you get any error messages when doing that? Try rebooting the computer too if you haven’t already.