Quickstart Guide¶
Download module from the marketplace¶
- go to DEEP Open Catalog
- Browse available modules
- Find the module and get it either from Docker Hub (easy) or Github (pro)
Run a module locally¶
Docker Hub way (easy)¶
Prerequisites
- docker
- If you want GPU support you can install nvidia-docker along with docker or install udocker instead of docker. udocker is entirely a user tool, i.e. it can be installed and used without any root priveledges, e.g. in a user environment at HPC cluster.
- Run the container
To run the Docker container directly from Docker Hub and start using the API simply run the following:
Via docker command:
$ docker run -ti -p 5000:5000 -p 6006:6006 deephdc/deep-oc-module_of_interestWith GPU support:
$ nvidia-docker run -ti -p 5000:5000 -p 6006:6006 deephdc/deep-oc-module_of_interestVia udocker:
$ udocker run -p 5000:5000 -p 6006:6006 deephdc/deep-oc-module_of_interestVia udocker with GPU support:
$ udocker pull deephdc/deep-oc-module_of_interest $ udocker create --name=module_of_interest deephdc/deep-oc-module_of_interest $ udocker setup --nvidia module_of_interest $ udocker run -p 5000:5000 -p 6006:6006 module_of_interest
- Access the module via API
To access the downloaded module via API, direct your web browser to http://127.0.0.1:5000. If you are training a model, you can go to http://127.0.0.1:6006 to monitor the training progress (if such monitoring is available for the model).
For more details on particular models, please, read model documentation.
Github way (pro)¶
Prerequisites
Using Github way allows to modify the Dockerfile for including additional packages, for example.
Clone the DEEP-OC-module_of_interest github repository:
$ git clone https://github.com/indigo-dc/DEEP-OC-module_of_interest
Build the container:
$ cd DEEP-OC-module_of_interest $ docker build -t deephdc/deep-oc-module_of_interest .
Run the container and access the module via API as described above
Note
One can also clone the source code of the module, usually located in the ‘module_of_interest’ repository.
Run a module on DEEP Pilot Infrastructure¶
Prerequisites
- DEEP-IAM registration
- oidc-agent installed and configured for DEEP-IAM (see rclone howto).
- orchent tool
If your are going to use DEEP-Nextcloud for storing you data you also have to:
- Register at DEEP-Nextcloud
- Include rclone installation in your Dockerfile (see rclone howto)
- Include call to rclone in your code (see rclone howto)
In order to submit your job to DEEP Pilot Infrastructure one has to create TOSCA YAML file.
The submission is then done via
$ orchent depcreate ./topology-orchent.yml '{}'
If you also want to access DEEP-Nextcloud from your container via rclone, you can create a following bash script for job submission:
#!/bin/bash
orchent depcreate ./topology-orchent.yml '{ "rclone_url": "https://nc.deep-hybrid-datacloud.eu/remote.php/webdav/",
"rclone_vendor": "nextcloud",
"rclone_user": <your_nextcloud_username>
"rclone_pass": <your_nextcloud_password> }'
To check status of your job
$ orchent depshow <Deployment ID>
Integrate your model with the API¶

The DEEPaaS API enables a user friendly interaction with the underlying Deep Learning modules and can be used both for training models and doing inference with the services. Check the full API guide for the detailed info.
The integration with the API is based on the definition of entrypoints to the model and the creation of standard API methods (eg. train, predict, etc). An easy way to integrate your model with the API and create Dockerfiles for building the Docker image is to use our DEEP DS template when developing your model.