Skip to content

Latest commit

 

History

History
40 lines (34 loc) · 5.47 KB

FORMAT.md

File metadata and controls

40 lines (34 loc) · 5.47 KB
Raw

Format of YAML Configuration Files for DL4MicEverywhere

DL4MicEverywhere follows the standards of the BioImage Model Zoo.

Each notebook is described using a configuration.yaml file with the metadata of the notebook as well as the parameters required for Docker container creation. A collection of configuration files for different versioned notebooks is already available. This collection is also versioned and tested through continuous integration pipelines. The users have also the flexibility to create and use their own configurations locally for their notebooks.

The structure of these YAML file is the same as in the BioImage Model Zoo with an additional field inside the config field:

  • config:
    • dl4miceveryhwere:
      • notebook_url: The URL to the Jupyter notebook that is stored in the Docker image.
      • requirements_url: The URL Python dependencies that are installed in the Docker image.
      • cuda_version: The CUDA version in the Docker image. The CUDA version is used to pull the Ubuntu image from Docker Hub (from Nvidia if GPU is selected). Its version (and the ubuntu_version) should be one from the following list.
      • cudnn_version: The cuDNN (CUDA Deep Neural Network) library's version in the Docker image. The cuDNN version is used for the installation of libraries that will allow the training of Deep Neural Networks(if GPU is selected). The version of cuDNN is related with the CUDA version, therefore it will need to be a specific version that is determined by the libcudnn library in Ubuntu 20.04 or Ubuntu 22.04.
      • ubuntu_version: The Ubuntu version for the operating system in the Docker image. This version is used to pull the base image from Docker Hub (from Nvidia if GPU is selected and from Ubuntu otherwise). The selected Ubuntu version must be between among the supported tags.
      • python_version: The installed Python version in the docker image.
      • sections_to_remove: The sections from the ZeroCostDL4Mic style notebook that are removed. Must be separated by spaces and follow the X.X. format. To keep all the sections, leave this argument empty; you still need to declare it.
      • description: A single-line description of the notebook that is shown in the GUI.
      • notebook_version: The version of the notebook in notebook_url. For ZeroCostDL4Mic notebooks the version is taken from the ZeroCostDL4Mic GitHub repository.
      • [optional] dl4miceverywhere_version: The DL4MicEverywhere version. Normally, this argument will be automatically generated when the configuration.yaml file is added or modified. The version is taken from the construct.yaml file. But in case you want to create and use your own configuration, this attribute is not needed (it can impact the tag of the Docker image).
      • [optional] docker_hub_image: The name of the created Docker image. This name is used to check if there is an existing image on the DockerHub repository from henriqueslab. Normally, this argument ise automatically generated when the configuration.yaml file is added or modified. The value is a combination of the notebook name, the notebook_version and the dl4miceverywhere_version (e.g. care_2d_zerocostdl4mic-z1.15.2-d1.0.0, being care_2d_zerocostdl4mic the name of the notebook, z indicating that is a ZeroCostDL4MicEverywhere notebook, 1.15.2 is the version of the notebook and 1.0.0 the version of the DL4MicEverywhere repository). If you want to create and use your configuration, this attribute is not needed. The tag will be generated from the available information on the configuration.yaml file.

As an example, here is the specific DL4MicEverywhere config field for the configuration.yaml file of CARE 2D notebook:

config:
  dl4miceverywhere:
    notebook_url: https://raw.githubusercontent.com/HenriquesLab/ZeroCostDL4Mic/master/Colab_notebooks/CARE_2D_ZeroCostDL4Mic.ipynb
    requirements_url: https://raw.githubusercontent.com/HenriquesLab/ZeroCostDL4Mic/master/requirements_files/CARE_2D_requirements_simple.txt
    cuda_version: 11.8.0
    cuda_version:  8.6.0.163
    ubuntu_version: '22.04'
    python_version: '3.10'
    sections_to_remove: 1.1. 1.2. 2. 6.3.
    description: CARE_2D_DL4Mic is the conversion of the 2D CARE from ZeroCostDL4Mic. 
    notebook_version: 1.15.2
    dl4miceverywhere_version: 1.0.0
    docker_hub_image: care_2d_zerocostdl4mic-z1.15.2-d1.0.0

Find the complete file here.