Environment setup

Conda, the AWS CLI, and the S3 profile are set up once for all of MAGiC, and are documented in CARB PEcAn Environment Setup. Follow that first.

This page covers the steps specific to the uncertainty workflow.

Clone the repository

git clone https://github.com/ccmmf/uncertainty
cd uncertainty

The rest of the instructions assume that you are working inside the uncertainty repository directory.

Every time you start a new terminal session

Return to the uncertainty repository directory and activate conda before continuing.

cd /path/to/uncertainty
conda activate ~/.conda/envs/pecan-all/
export AWS_PROFILE=magic

SIPNET

The analyses run the SIPNET binary that the ensemble workflow builds. After a magic-ensemble prepare step, it is at sipnet.git in that run directory; the configuration file points to it.

The configuration file

The uncertainty CLI uses a configuration file to control the workflow. For the demo, the configuration file is provided with the demo data. For your own runs, copy and edit example_user_config.yaml.

Confirm setup

./magic-uncertainty help            # CLI runs
aws s3 ls s3://carb/pfts/           # bucket reads

If either fails, fix it before starting the demo.

The model runs and the steps that read their output should run on a compute node.

---

Next: Uncertainty demo.

Overview: MAGiC uncertainty training.