Skip to content

dylanmartin/verbose-winner

Repository files navigation

Running testApp in Flare simulator

  • Launch the docker container using ./dockerRun.sh
  • Run the following command inside the container: nvflare simulator ./testApp/ -w /tmp/nvflare/workspace_folder/ -t 2

Running a job in POC mode

  • Launch the docker container using ./dockerRun.sh
  • Run the following commands inside the container:
    • export NVFLARE_POC_WORKSPACE="/my-workspace/poc-workspace"
    • nvflare poc --prepare
    • nvflare poc --start
  • This will put the terminal in the Flare Console
  • Put custom apps in my-workspace/poc-workspace/admin/transfer
  • submit_job <job_folder> will run the job at admin/transfer/<job_folder>

Run the production-mode-test

  • cd into production-mode-test/
  • follow instructions in the production-mode-test/README.md

Run an app that uses local code

  • In the container, set the python path to your local code directory. EX export PYTHONPATH=$PYTHONPATH:/my-workspace/poc-workspace/localcode
  • In the config files for the app, change the component paths to point to the directory containing your modules.
  • Example:
    • The localcode/ directory contains a folder called np which is a python module with __init__.py and np_trainer.py
    • config_fed_client.json for the app looks like this:
{
    "format_version": 2,
    "executors": [
      {
        "tasks": [
          "train"
        ],
        "executor": {
          "path": "np.np_trainer.NPTrainer",
          "args": {}
        }
      }
    ],
    "task_result_filters": [],
    "task_data_filters": [],
    "components": []
}

About

Experimenting with running a federated computation using NVIDIA Flare

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages