- Launch the docker container using
./dockerRun.sh - Run the following command inside the container:
nvflare simulator ./testApp/ -w /tmp/nvflare/workspace_folder/ -t 2
- Launch the docker container using
./dockerRun.sh - Run the following commands inside the container:
export NVFLARE_POC_WORKSPACE="/my-workspace/poc-workspace"nvflare poc --preparenvflare 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 atadmin/transfer/<job_folder>
- cd into
production-mode-test/ - follow instructions in the
production-mode-test/README.md
- 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 callednpwhich is a python module with__init__.pyandnp_trainer.py - config_fed_client.json for the app looks like this:
- The
{
"format_version": 2,
"executors": [
{
"tasks": [
"train"
],
"executor": {
"path": "np.np_trainer.NPTrainer",
"args": {}
}
}
],
"task_result_filters": [],
"task_data_filters": [],
"components": []
}