1. CD into a directory where pipelines will be stored, e.g.

mkdir -p ~/.clara/pipelines
cd ~/.clara/pipelines

2. Use clara pull pipeline to download the pipeline definition

clara pull pipeline clara_multiai_pipeline

3. Unzip test data

unzip app_multiai-input_v1.zip

4. Unzip and Install models

sudo unzip app_multiai-model_v1.zip 
sudo cp -r multiai-models/* /clara/common/models/

4. Workaround for 0.5.0-preview-1 release - pipeline definitions in the 0.5.0-preview-1 release does not contain full path to nvcr.io and does not include default pull secret. Edit the pipeline definition file and add the following after

name:
pull-secrets:
  - ngc-clara

Find and prepend all image tags that start with clara/ with nvcr.io/nvstaging/

For example, change

name: my-pipeline
operators:
  - name: clara-hqs
    container:
      image: clara/clara-hqs

to

name: my-pipeline
pull-secrets:
  - ngc-clara
operators:
  - name: clara-hqs
    container:
      image: nvcr.io/nvidia/clara/clara-hqs

6. Install pipeline

clara create pipelines -p multiAI-mhd-pipeline.yaml

Store the pipeline id. 7. Verify that the pipeline is correctly stored

clara list pipelines

8. Verify that the stored pipeline is correct by examining the output of the following command

kubectl  describe pipeline <pipeline ID>

9. Prepare the data as described in overview page in "Data and Models" section. Store the data path.
10. Create the job

clara create jobs -n <job_name> -p <pipeline_id> -f <data_path>

Store the job id after executing the above command.

11. Start the job

clara start job -j <job_id>