GenSynth Documentation

Configuration

The file config.sh sets a number of shell variables used by the start.sh script. These MUST be configured before running the start script.

Change these variables in config.sh if desired. Note that any paths should be absolute, or relative to this directory containing the start.sh script.

Normally you only need to change these parameters.

GENSYNTH_GPU_RESOURCES: A list of available GPU resources. In a multi-machine setup, the number of available GPUs on each machine must be specified using a space-separated-value list of <machine-hostname>:<number-of-gpus>. The default is 8 GPUs available on the API server machine.

GENSYNTH_API_HOSTNAME: Typically you have to change GENSYNTH_API_HOSTNAME to the name of the server so that your browser can locate the API. The default of localhost is only acceptable if you are running GenSynth on your workstation or access the API by ssh-tunneling of ports from your localhost.

GENSYNTH_WEB_PORT: 80 by default, the port for the web server to listen on.

GENSYNTH_PACKAGE_REPO: The folder containing the packages to be installed, ./gensynth-packages by default, mounted to /gensynth/packages in the API container.

GENSYNTH_WORKSPACE_PATH: The folder for inputs, outputs, and working space used by GenSynth, ./gensynth-workspace by default, mounted to /gensynth/workspace in the container. This should be a large disk; you may want to use a network mount.

  • Data and input models may be placed here (as done by examples and tutorials)

  • This is where models and other data will be written by GenSynth

GENSYNTH_EXTRA_PIP_REQUIREMENTS_PATH: A pip requirements file of packages to install on the GenSynth server, within the Docker container, defaults to pip_requirements.txt The format of this file should be suitable for use with pip install -r

GENSYNTH_LICENSE_ACTIVATION_METHOD: Defaulting to online, this may be set to offline if GenSynth does not have internet access.