Run locally

[g30ba1]

I've built all the containers successfully.

But I've NO idea how to run them.

Any tip?

[dusty-nv]

Hi @g30ba1 , you can find the run instructions on NGC. For example:

https://ngc.nvidia.com/catalog/containers/nvidia:l4t-ml

sudo docker run -it --rm --runtime nvidia --network host nvcr.io/nvidia/l4t-ml:r32.4.2-py3

[uersoy]

Hi @g30ba1 , you can find the run instructions on NGC. For example:

https://ngc.nvidia.com/catalog/containers/nvidia:l4t-ml

sudo docker run -it --rm --runtime nvidia --network host nvcr.io/nvidia/l4t-ml:r32.4.2-py3

Hello @dusty-nv,

I can run the container at localhost:8888 however all my saved work gets lost the next time I access the same. Seems like no checkpoint is saved when I save my jupyterlab files. Do you know why this may be happening?

Thanks!

[g30ba1]

Hi @g30ba1 , you can find the run instructions on NGC. For example:

https://ngc.nvidia.com/catalog/containers/nvidia:l4t-ml

sudo docker run -it --rm --runtime nvidia --network host nvcr.io/nvidia/l4t-ml:r32.4.2-py3

Hi Dusty, thank you for your time.

I found that if we re-built the containers, the command to run is:

sudo docker run -it --rm --runtime nvidia --network host -v /home/user/Documents/Projects:/home l4t-tensorflow:r32.4.2-tf1.15-py3

(-v flag indicates that we will be working on a local folder)

NGC containers are a nice tool to get an environment as quickly as possible.

[g30ba1]

Hi @g30ba1 , you can find the run instructions on NGC. For example:
https://ngc.nvidia.com/catalog/containers/nvidia:l4t-ml

sudo docker run -it --rm --runtime nvidia --network host nvcr.io/nvidia/l4t-ml:r32.4.2-py3

Hello @dusty-nv,

I can run the container at localhost:8888 however all my saved work gets lost the next time I access the same. Seems like no checkpoint is saved when I save my jupyterlab files. Do you know why this may be happening?

Thanks!

You must use the -v flag to WORK on a local folder while using the container, i.e.:

sudo docker run -it --rm --runtime nvidia --network host -v /home/user/Documents/Projects:/home l4t-tensorflow:r32.4.2-tf1.15-py3

If you are suing the container published on Nvidia´s NGC, the command is:

sudo docker run -it --rm --runtime nvidia --network host -v /home/user/project:/location/in/container nvcr.io/nvidia/l4t-ml:r32.4.2-py3

To use a local folder, you can use any path existent on your local machine (host), or you can create the folder while you´re INSIDE the container.

[uersoy]

Hi @g30ba1 , you can find the run instructions on NGC. For example:
https://ngc.nvidia.com/catalog/containers/nvidia:l4t-ml

sudo docker run -it --rm --runtime nvidia --network host nvcr.io/nvidia/l4t-ml:r32.4.2-py3

Hello @dusty-nv,
I can run the container at localhost:8888 however all my saved work gets lost the next time I access the same. Seems like no checkpoint is saved when I save my jupyterlab files. Do you know why this may be happening?
Thanks!

You must use the -v flag to WORK on a local folder while using the container, i.e.:

sudo docker run -it --rm --runtime nvidia --network host -v /home/user/Documents/Projects:/home l4t-tensorflow:r32.4.2-tf1.15-py3

If you are suing the container published on Nvidia´s NGC, the command is:

sudo docker run -it --rm --runtime nvidia --network host -v /home/user/project:/location/in/container nvcr.io/nvidia/l4t-ml:r32.4.2-py3

To use a local folder, you can use any path existent on your local machine (host), or you can create the folder while you´re INSIDE the container.

Thank you Jorge. I can use a local folder to save my scripts, notebooks however I am wondering how to keep the changes I make to the whole container image. Let's say I install new dependencies, libraries on top of existing default ones. Will I have to create a new container image?

[uersoy]

Hi @g30ba1 , you can find the run instructions on NGC. For example:
https://ngc.nvidia.com/catalog/containers/nvidia:l4t-ml

sudo docker run -it --rm --runtime nvidia --network host nvcr.io/nvidia/l4t-ml:r32.4.2-py3

Hello @dusty-nv,
I can run the container at localhost:8888 however all my saved work gets lost the next time I access the same. Seems like no checkpoint is saved when I save my jupyterlab files. Do you know why this may be happening?
Thanks!

You must use the -v flag to WORK on a local folder while using the container, i.e.:
sudo docker run -it --rm --runtime nvidia --network host -v /home/user/Documents/Projects:/home l4t-tensorflow:r32.4.2-tf1.15-py3
If you are suing the container published on Nvidia´s NGC, the command is:
sudo docker run -it --rm --runtime nvidia --network host -v /home/user/project:/location/in/container nvcr.io/nvidia/l4t-ml:r32.4.2-py3
To use a local folder, you can use any path existent on your local machine (host), or you can create the folder while you´re INSIDE the container.

Thank you Jorge. I can use a local folder to save my scripts, notebooks however I am wondering how to keep the changes I make to the whole container image. Let's say I install new dependencies, libraries on top of existing default ones. Will I have to create a new container image?

I figured out. The below link explains it.
https://docs.nvidia.com/deeplearning/frameworks/user-guide/index.html

[g30ba1]

Nice find! I'm already reading it.

Regarding your question, yes, you have to re-built an image, to add libraries, dependencies or any other requirements for your projects.

[uersoy]

Nice find! I'm already reading it.

Regarding your question, yes, you have to re-built an image, to add libraries, dependencies or any other requirements for your projects.

I also read the whole thing to learn but I found the answer to my question in Section 10.1.4.