Use cases - in prioritised order

• Can login to MLReef

  Users can log into MLReef with their own credentials and only have access to their own repositories as well as public repositories.

  exkl. account creation & management

  • backend#8 Gitlab token as MLReef token
  • frontend#140 Login in Frontend via Gitlab/Backend

• Can create and fork projects

  • backend#16 jpa
  • frontend#99
  • frontend#157 Create new Project

• Can push and pull their data via Git

  Our Alpha users shall be able to leverage the full feature set of Git for their repositories. This includes uploading of large files. Thus we have to provide our self hosted Git infrastructure (via Gitlab CE)

  • backend#2 Setup Gitlab Instance
  • backend#6 Bug in runner naming

• Can execute built in algorithm experiments and get results

  Make sure that our users can configure and execute the built in algorithms and are able to download the model binary from the GUI

  • epf#45 GPU Processing
  • epf#48
  • frontend#161
  • backend#15

• Can review the experiment statistics

  Users shall be able to look at the metadata generated by the machine training. For this the epf has to persist the training configuration, statistics, etc. to the backend which then in turn makes those data available to the frontend.

  • frontend#153 Expand button in Experiment Cards should always work
  • frontend#154 Show experiment information
  • frontend#148 Combine same experiment cards

• Can view their data

  Users can navigate through the repository and look at all their data, including their active Data Instances.

  • frontend#111 show commit data in file view
  • frontend#114 show commit avatars in file view
  • frontend#37

• Can Perform basic editing functions online (1)

  • frontend#160 Delete File
  • frontend#159 Create new empty file
  • Edit Textfile

• Can Execute Pipelines on GPU Instances

  Training on CPUs is much too slow for what we want to achieve at MLReef, so the EPF has to be able to harness the GPU computing power of GPU instances.

  • epf#45 Make dockerfile GPU compatible for AWS integration

• Can execute built in Data Operations

  Users can select from the currently available Data Operations, add them to a Data Pipeline and execute the Data Pipeline.

  • [frontend#83 multiply data operation
  • frontend#119 data instance cards

• Can Execute custom Data Operation and Model Algorithm

  • epf#34 EPF Battle Plan
  • epf#31 define data processing
  • epf#46 implement decorators
  • sar-esa#4 Naming Convention for new Files

• Can Load previous execution configurations Data Scientists shall be able to repeat a previous experiment.]

  • frontend#150 Load previous experiment configuration
  • frontend#164 Load previous data operation configuration

• Can view and explorer Data Instance

• Can View Algorithm Execution summary (and statistics)

  • frontend#147 Diff view with images

• Can View Algorithm Execution details

• Can execute and view data visualisations

  Users can select from the built in Data Visualisations, execute them and view the resulting visualisation in the GUI.

  • frontend#155
  • frontend#68 Viz. Pipeline

• Can Save and Load previous Visualisation execution configuration

• Misc. GUI Improvements

  • frontend#146 Limit Commit message to GUI size
  • frontend#151 validate pipeline input earlier
  • frontend#149 Get user group name

Stretch Goals

Can Perform basic editing functions online (2)

*frontend#145 create new branch

• Online IDE

• Merge Request

• frontend#158 Show merge Request bar

• Images in Markdown View

• frontend#152 GUI filtering

Projects (1 project : n repositories)

User management via the MLReef GUI

Additional Notes