Welcome to the Transfer Learning Toolkit Forum

Feel free to post any questions or thoughts about the Nvidia Transfer Learning Toolkit. Nvidia staff are here to assist you.

Hi, I couldn’t find the transfer learning toolkit within the NVIDIA GPU CLOUD, under CATALOG. How or where can I access it?

Thanks

Hi,
In Nvidia GPU Cloud, you can navigate to ‘Container’, you should be able to see the Toolkit there

Thanks

Hi,

Thanks for the quick response. But unfortunately the list under “ALL CONTAINERS” is empty.

I received the email regards the E/A for TLT. I registered an account for NGC. Did I miss any of the steps required?

Thanks

Hi,

Can you access the “Getting Started guide”? Can you pull docker using “docker pull” command given there?

Thanks

Hi madil,

Thanks for replying. I got the access, able to pull the docker as well.

Can I know which “Getting Started Guide” you are referring to?

Hi cw.yong,

You can find the “Getting Started guide” for download in the Early access Member area.

Adil

Hi madil,

Thank you for the info.

How do I find ‘Transfer Learning Toolkit’ at ngc cloud?
I can not see it in the container list.
Will it not be listed without the E/A?
Thanks.

Dear

I have already applied for TLT for one month, which is not approved. Could you please help?

Thanks

Hanling zhang

Hunan university

My gpu is 940mx
And cuda can support 940m
in this case, cant’t I get support for my gpu?
940mx is upgrade version of 940m

Here is the url
https://ngc.nvidia.com/containers/nvtltea:iva:tlt-streamanalytics

You need to select org = nvtltea and team = iva when you are login into https://ngc.nvidia.com/

thank you for your response, I have managed to find the tool. However, the org and team differentiation puzzles me. Is it a mean to share a tool to a colleague who is a devtalk developer or just some internal intrinsic design?

Thanks for the direct link. It save me a great deal of headache.

Hello, I want to use Transfer Learning Toolkit to train my own dataset by using Deepstream. I want to try it on my Jetson Nano Jetpack 4.2.2. The problem is how to install nvidia-docker2 and also NVIDIA GPU drivers on Jetson Nano?
Nvidia-docker2 and NVIDIA GPU drivers are the requirements to deploy transfer learning toolkit.

Nvidia docker can be installed if you are using the latest Jetpack: “JetPack 4.2.1 also introduced two beta features: NVIDIA Container Runtime with Docker integration and TensorRT support for INT-8 DLA operations.” [ source https://developer.nvidia.com/embedded/jetpack ]
The installation of the Nvidia docker will proceed via the default Jetpack 4.2.2 interface. You will see the docker listed among other components.
However, the requirement from the documentation: “NVIDIA GPU driver v410.xx or above” implies that x86_64 Computer will be used for running the toolkit, as it seems to me.

Thank you Andrey1984 for the reply, that means I just need Nvidia Docker 2 to deploy Transfer Learning Toolkit right? After I Installed the nvidia-docker2, I got some error like ‘error from daemon’, do you have any idea?

Are you installing nvidia-docker2 at x86_64 PC? at Jetson nano device? somewhere else?
It seems that you might need to stick to use of x86_64 architecture.
Refer to the corresponding documentation, please:
https://docs.nvidia.com/metropolis/TLT/tlt-getting-started-guide/index.html
https://ngc.nvidia.com/catalog/containers/nvidia:tlt-streamanalytics
https://medium.com/dataseries/build-and-deploy-accurate-deep-learning-models-for-intelligent-image-and-video-analytics-8ad755213c06
https://docs.nvidia.com/metropolis/TLT/tlt-release-notes/index.html

Thank you Andrey1984 for the advice and assists, currently I am trying to run transfer learning toolkit at Jetson Nano decive with Jetpack 4.2.2, and I still need nvidia-docker2 to run Transfer Learning toolkit right? Or nvidia-docker2 is for x86_64 PC too?

the toolkit is for x86_64;
for installing docker[nvidia-docker2] at x86_64 you may use https://github.com/NVIDIA/nvidia-docker
for nano you may use tlt-converter: reference thread