Updating tao toolkit

[david9xqqb]

Please provide the following information when requesting support.
• Hardware RTX3090
• Network Type unet / vgg16
• TLT Version

dockers: [‘nvidia/tao/tao-too lkit-tf’, ‘nvidia/tao/tao-toolkit-pyt’, ‘nvidia/tao/tao-toolkit-lm’]
format_version: 2.0
toolkit_version: 3.22.02
published_date: 02/28/2022

What’s the recommended procedure to assure we are using the latest version of the docker images for tao?

Thanks!!

[Morganh]

You can check the info of “$tao info --verbose”

[david9xqqb]

Yes I know. In fact, the output of that is in the post. My question s what is the procedure to update the dockers.

For example my environment has:

Running command in container: nvcr.io/nvidia/tao/tao-toolkit-tf:v3.21.11-tf1.15.5-py3

But the latest tag is v3.22.05-tf1.15.4-py3

[Morganh]

Please note that nvcr.io/nvidia/tao/tao-toolkit-tf:v3.21.11-tf1.15.5-py3 and nvcr.io/nvidia/tao/tao-toolkit-tf:v3.21.11-tf1.15.4-py3 are one of the 3.21.11 versions.

The 15.4 is for detectnet_v2 and faster_rcnn.
The 15.5 is for other object detection models.

[david9xqqb]

No real solution yet

I thought this would solve

docker pull nvcr.io/nvidia/tao/tao-toolkit-tf:v3.22.05-tf1.15.4-py3

But It doesn’t get loaded next time I run tao

So the question remains:

What’s the recommended procedure to assure we are using the latest version of the docker images for tao?

[Morganh]

If you trigger TAO via tao-launcher, please install the latest nvidia-tao CLI package.
If you trigger TAO via "docker run xxx ", you can “docker pull xxx” the latest tao docker directly and then run "docker run xxx ". You can pull the latest from TAO Toolkit for Computer Vision | NVIDIA NGC.

[david9xqqb]

Running

pip3 install --upgrade nvidia-tao

Does not upgrade:

(TAO) david@AI01:~$ tao info
Configuration of the TAO Toolkit Instance
dockers: [‘nvidia/tao/tao-toolkit-tf’, ‘nvidia/tao/tao-toolkit-pyt’, ‘nvidia/tao/tao-toolkit-lm’]
format_version: 2.0
toolkit_version: 3.22.02
published_date: 02/28/2022
(TAO) david@AI01:~$ pip3 install --upgrade nvidia-tao
Looking in indexes: Simple index, https://pypi.ngc.nvidia.com
Requirement already satisfied: nvidia-tao in ./Envs/TAO/lib/python3.8/site-packages (0.1.21)
Requirement already satisfied: six==1.15.0 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (1.15.0)
Requirement already satisfied: requests==2.24.0 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (2.24.0)
Requirement already satisfied: tabulate==0.8.7 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (0.8.7)
Requirement already satisfied: idna==2.10 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (2.10)
Requirement already satisfied: urllib3==1.25.10 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (1.25.10)
Requirement already satisfied: certifi==2020.6.20 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (2020.6.20)
Requirement already satisfied: websocket-client==0.57.0 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (0.57.0)
Requirement already satisfied: docker-pycreds==0.4.0 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (0.4.0)
Requirement already satisfied: docker==4.3.1 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (4.3.1)
Requirement already satisfied: chardet==3.0.4 in ./Envs/TAO/lib/python3.8/site-packages (from nvidia-tao) (3.0.4)
WARNING: You are using pip version 22.0.4; however, version 22.1.1 is available.
You should consider upgrading via the ‘/home/david/Envs/TAO/bin/python -m pip install --upgrade pip’ command.
(TAO) david@AI01:~$ tao info
Configuration of the TAO Toolkit Instance
dockers: [‘nvidia/tao/tao-toolkit-tf’, ‘nvidia/tao/tao-toolkit-pyt’, ‘nvidia/tao/tao-toolkit-lm’]
format_version: 2.0
toolkit_version: 3.22.02
published_date: 02/28/2022
(TAO) david@AI01:~$

[Morganh]

Do you mean you want to upgrade to 22.05 version?

[david9xqqb]

Yes, but more so, I want the procedure to keep up to date.

(Quickest reply ever! thanks!!!)

[Morganh]

The latest 22.05 is not announced. Its corresponding tao-launcher is not ready yet.
Please stay tuned. Thanks a lot.