Run a job with Jupyter

1. Pull a docker image with jupyter

Let's use the default tensorflow image:

$ docker pull tensorflow/tensorflow:latest-gpu

2. Upload the image

$ neuro image push tensorflow/tensorflow:latest-gpu

3. Create a job

$ neuro job submit --http 8888 --memory 2G --cpu 1 registry.staging.neuromation.io/<USERNAME>/tensorflow/tensorflow:latest-gpu

Job ID: job-3cd3d725-8181-4bd6-8360-776a043a889c Status: pending
Shortcuts:
  neuro job status job-3cd3d725-8181-4bd6-8360-776a043a889c  # check job status
  neuro job monitor job-3cd3d725-8181-4bd6-8360-776a043a889c # monitor job stdout
  neuro job kill job-3cd3d725-8181-4bd6-8360-776a043a889c    # kill job

Watch neuro job status job-3cd3d725-8181-4bd6-8360-776a043a889c untill you get status running:

Job: job-3cd3d725-8181-4bd6-8360-776a043a889c
Owner: rebryk
Status: running
Image: registry.staging.neuromation.io/rebryk/tensorflow/tensorflow:latest-gpu
Command: None
Resources: Resources(memory=2048, cpu=1.0, gpu=None, shm=False, gpu_model=None)
Http URL: https://job-3cd3d725-8181-4bd6-8360-776a043a889c.jobs.platform.staging.neuromation.io
Created: 2019-01-13T11:52:20.893132+00:00
Started: 2019-01-13T11:54:02.729816+00:00

4. Open Jupyter

Follow the link from the status. In this case https://job-3cd3d725-8181-4bd6-8360-776a043a889c.jobs.platform.staging.neuromation.io:

As you can see, you need an access token to login into your notebook. Here is a way to get it:

$ neuro job monitor job-3cd3d725-8181-4bd6-8360-776a043a889c

...
[I 11:54:06.367 NotebookApp] The Jupyter Notebook is running at:
[I 11:54:06.367 NotebookApp] http://(job-3cd3d725-8181-4bd6-8360-776a043a889c or 127.0.0.1):8888/?token=a4a39a2b676a11a3f0574d59fdedfa47d53dc95bf6993fc4
...

So we should use a4a39a2b676a11a3f0574d59fdedfa47d53dc95bf6993fc4.

Enjoy!