Questions about how to proceed Tao Toolkit

[hitesarang1]

I want to know if this pipeline I understand is correct to start TAO Toolkit Quick Start. I’m confused whether I should build a docker container first and run a Jupyter Notebook there, or run Jupyter Notebook directly in a virtual environment, as in the introduction video. I already downloaded all the prerequisites.

• Like the introduction video

1. Turn on the terminal and activate virtual env(miniconda)
2. docker login nvcr.io
3. cd <path ~cv_samples_v1.3.0>
4. Deploy jupyter notebook
   (There is nothing on $docker ps -a)
   (In this way, I can’t understand how the local folder mounts to /workspace path)

Or

• Use Docker Container

1. Turn on the terminal and activate virtual env(miniconda)
2. docker login nvcr.io
3. docker run --runtime=nvidia --name taotool -it -v /home/desktop/tao_tool/cv_samples_v1.3.0:/workspace -p 8888:8888 nvcr.io/nvidia/tao/tao-toolkit-tf:v3.21.11-tf1.15.5-py3 /bin/bash
   (docker exec -it taotool /bin/bash)
4. jupyter notebook --ip 0.0.0.0 --port 8888 --allow-root
   (Does I use a virtual environment in this case?)

Is both ways possible??

[Morganh]

@hitesarang1

I should build a docker container first

There are already TAO container. It is not needed for end user to build any docker.

@hitesarang1

cd <path ~cv_samples_v1.3.0>

TAO is already updated to 4.0 as of now. The latest notebook is here. TAO Toolkit Quick Start Guide - NVIDIA Docs . Maybe you are watching an old introduction video from 3rd party.

Suggest you to run TAO with installing tao launcher. See more info in TAO Toolkit Quick Start Guide - NVIDIA Docs.

@hitesarang1

Does I use a virtual environment in this case?

In this case, it is not needed. See TAO Toolkit Quick Start Guide - NVIDIA Docs

[Morganh]

Please generate a new topic since it is different from original question. Thanks.