AssertionError: Config path must be a valid unix path. No file found at: /root/.docker/config.json. Did you run docker login?

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Please provide the following information when requesting support.

• Hardware (T4/V100/Xavier/Nano/etc)
• Network Type (Detectnet_v2/Faster_rcnn/Yolo_v4/LPRnet/Mask_rcnn/Classification/etc) Classification_tf2
• TLT Version (Please run “tlt info --verbose” and share “docker_tag” here) 4.0.0-tf2.9.1
• Training spec file(If have, please share here) Default
• How to reproduce the issue ? (This is for errors. Please share the command line and the detailed log here.)

Initial setup using
wget --content-disposition https://api.ngc.nvidia.com/v2/resources/nvidia/tao/tao-getting-started/versions/4.0.1/zip -O getting_started_v4.0.1.zip

unzip -u getting_started_v4.0.1.zip -d ./getting_started_v4.0.1 && rm -rf getting_started_v4.0.1.zip && cd ./getting_started_v4.0.1

docker login nvcr.io done with API key.

Then to run the Jupyter notebook, launched a container with docker run -it --gpus all --name amogh_tao -p 8888:8888 -v /$PWD:/$PWD -v /var/run/docker.sock:/var/run/docker.sock -w /$PWD nvcr.io/nvidia/tao/tao-toolkit:4.0.0-tf2.9.1 bash and launched Jupyter from within the container.

The classification_tf2 notebook installs NGC CLI and downloads the pre-trained model.

When I try to train from within the notebook, it says API key missing.

Looked up AssertionError: Config path must be a valid unix path. No file found at: /root/.docker/config.json. Did you run docker login? but that didn’t help

1 Like
2

Could you share the full command and full log? Thanks.

3

I have another issue related to that actually that popped up
While I was trying to fix this, I decided to try to log in to docker as user instead of root.
Since then, when I run the classification_tf2 notebook, the train command, it shows something like the following.

image
image1920×500 135 KB

4

In the notebook when I try running the !tao train … command, it says that tao: command not found.

5

Suggest you to double check with the guide in TAO Toolkit Quick Start Guide - NVIDIA Docs, especially

  • setting up a python environment using miniconda
  • Install the CLI launcher via the quick start script
  • trigger notebook
  • etc
6

Thanks
I was able to access the API key and log in I think

But since I am running the notebook and training command inside a docker, there’s another error.

Setup: Following is the command for starting my base docker container

image
image1920×357 48.2 KB

Once inside the container, I open the jupyter notebook classification_tf2.ipynb and run the cells.

When I run the training command -
!tao classification_tf2 train -e $SPECS_DIR/spec.yaml
I get the following error -

image
image2142×258 25.3 KB

I think when the training script tries launching a docker container while already being inside a docker container, there is this issue.

I read your comment and have followed the advice while running my docker container. Can you suggest what else can be done?

7

You are already running inside the container. In this case, please use command without “tao”.
i.e.,
! classification_tf2 train -e $SPECS_DIR/spec.yaml

BTW, you can also open a terminal instead of the jupyter notebook. And then run above command.

8

Thanks

9

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.