Tensorboard visualization


I try to “Tensorboard visualization” at https://docs.nvidia.com/clara/tlt-mi/tlt-mi-getting-started/index.html#unique_893928591.

I issued these commands.

root@80b04d9a6b65:/workspace/test/Nvidia-train-SDK/MODEL/segmentation_ct_lung_tumor/files/commands# python3 -m tensorboard.main --logdir=…/models/ &

root@80b04d9a6b65:/workspace/test/Nvidia-train-SDK/MODEL/segmentation_ct_lung_tumor/files/commands# TensorBoard 1.12.2 at http://80b04d9a6b65:6006 (Press CTRL+C to quit)

root@80b04d9a6b65:/workspace/test/Nvidia-train-SDK/MODEL/segmentation_ct_lung_tumor/files/commands# ./train_finetune.sh

But, I could not view “localhost:6006” or “http://80b04d9a6b65:6006”.

What should I do if I want to use Tensorboard ?

I think the problem could be with the ports. The port 6006 should be exposed when running the docker image in order to access tensorboard.

#docker image

docker run --runtime=nvidia -it --rm\
 -p 5000:5000 -p 6006:6006 \
 -v /your/local/clara/directory:/clara $dockerImage /bin/bash

Thank you!
I resolved this problem.

I commanded without -p 5000:5000 -p 6006:6006.


I am trying to solve same problem. I run the docker with command:
docker run --shm-size=1G --ulimit memlock=-1 --ulimit stack=67108864 --gpus=1 -p 6006:80 -it --rm -v /media/vavr/clara-experiments/:/workspace --ipc=host nvcr.io/nvidia/clara-train-sdk:v3.1.01 /bin/bash

I can run the server, I can ./train.sh.
The problem is that I cant visualize by tensorboard.
python3 -m tensorboard.main --logdir="/workspace/mmars/clara_ct_seg_spleen_no_amp_v1/models"

It writes TensorBoard 1.15.0+nv at http://d01e8da0183b:6006/
but this page is not found.

Please can you help me what I am doing wrong?
Thank you very much.

Thanks for you interest in clara train SDK. You have a couple of things wrong there

  1. wrong port mapping
  2. d01e8da0183b seems like the docker container id

The docker option -p maps <hostport external>: <internal container port>

so you in you case you mapped 6006 on you host machine to port 80 inside the docker
I don’t think you should use port 80 but if you insist then for this to work you should

  1. in th docekr run tensorboard --port 80 this will have tensorbaord use port 80 in the docker
  2. go to http: you machine ip: 6006 to see tensor board

If this doesn’t work please use other port then 80 may be 6000
hope that helps

Thank you