Federated Learning - fedml federate
fedml federate build [OPTIONS]
Options
Name | Default | Description |
---|---|---|
--dest_folder or -d | true | The destination package folder path. If this option is not specified, the built packages will be located in a subdirectory named fedml-federate-packages in the directory of YAML_FILE. |
Here is the command for building TensorOpera Federate packages.
At first, you need to define your job properties in the job yaml file, e.g., workspace, job entry commands.
This job yaml file can be from the launch job yaml file, which is as follows:
# Local directory where your source code resides.
# It should be the relative path to this job yaml file or the absolute path.
# If your job doesn't contain any source code, it can be empty.
workspace: .
# Running entry commands which will be executed as the job entry point.
# If an error occurs, you should exit with a non-zero code, e.g. exit 1.
# Otherwise, you should exit with a zero code, e.g. exit 0.
# Support multiple lines, which can not be empty.
job: |
echo "current job id: $FEDML_CURRENT_RUN_ID"
echo "current edge id: $FEDML_CURRENT_EDGE_ID"
echo "Hello, Here is the launch platform."
echo "Current directory is as follows."
pwd
echo "config file"
cat config/fedml_config.yaml
echo "current rank: $FEDML_CLIENT_RANK"
python3 torch_client.py --cf config/fedml_config.yaml --rank $FEDML_CLIENT_RANK --role client --run_id $FEDML_CURRENT_RUN_ID
# Running entry commands on the server side which will be executed as the job entry point.
# Support multiple lines, which can not be empty.
server_job: |
echo "Hello, Here is the server job."
echo "Current directory is as follows."
python3 torch_server.py --cf config/fedml_config.yaml --rank 0 --role server --run_id $FEDML_CURRENT_RUN_ID
job_type: federate # options: train, deploy, federate
# train subtype: general_training, single_machine_training, cluster_distributed_training, cross_cloud_training
# federate subtype: cross_silo, simulation, web, smart_phone
# deploy subtype: none
job_subtype: cross_silo
# Bootstrap shell commands which will be executed before running entry commands.
# Support multiple lines, which can be empty.
bootstrap: |
echo "Bootstrap finished."
computing:
minimum_num_gpus: 1 # minimum # of GPUs to provision
maximum_cost_per_hour: $3000 # max cost per hour for your job per gpu card
#allow_cross_cloud_resources: true # true, false
#device_type: CPU # options: GPU, CPU, hybrid
resource_type: A100-80G # e.g., A100-80G, please check the resource type list by "fedml show-resource-type" or visiting URL: https://TensorOpera.ai/accelerator_resource_type
data_args:
dataset_name: mnist
dataset_path: ./dataset
dataset_type: csv
model_args:
input_dim: '784'
model_cache_path: /Users/alexliang/fedml_models
model_name: lr
output_dim: '10'
training_params:
learning_rate: 0.004
The config items will be mapped to the equivalent environment variables with the following rules.
if the config path for config_sub_key is as follows.
config_parent_key:
config_sub_key: config_sub_key_value
Then the equivalent environment variable will be as follows.
FEDML_ENV_uppercase($config_parent_key)_uppercase($config_sub_key)
e.g., the equivalent environment variables of above example config items will be as follows.
dataset_name = $FEDML_ENV_DATA_ARGS_DATASET_NAME
dataset_path = $FFEDML_ENV_DATA_ARGS_DATASET_PATH
dataset_type = $FEDML_ENV_DATA_ARGS_DATASET_TYPE
model_name = $FEDML_ENV_MODEL_ARGS_MODEL_NAME
model_cache_path = $FEDML_ENV_MODEL_ARGS_MODEL_CACHE_PATH
input_dim = $FEDML_ENV_MODEL_ARGS_MODEL_INPUT_DIM
output_dim = $FEDML_ENV_MODEL_ARGS_MODEL_OUTPUT_DIM
Your may use these environment variables in your job commands. e.g.,
job: |
python3 torch_client.py --cf config/fedml_config.yaml --rank $FEDML_CLIENT_RANK --role client --run_id $FEDML_CURRENT_RUN_ID -m $FEDML_ENV_MODEL_ARGS_MODEL_NAME -mc $FEDML_ENV_MODEL_ARGS_MODEL_CACHE_PATH -mi $FEDML_ENV_MODEL_ARGS_MODEL_INPUT_DIM -mo $FEDML_ENV_MODEL_ARGS_MODEL_OUTPUT_DIM -dn $FEDML_ENV_DATA_ARGS_DATASET_NAME -dt $FEDML_ENV_DATA_ARGS_DATASET_TYPE -dp $FEDML_ENV_DATA_ARGS_DATASET_PATH
Examples
fedml federate build federated_job.yaml
Your client package file is located at: /home/fedml/launch/fedml-federate-packages/client-package.zip
Your server package file is located at: /home/fedml/launch/fedml-federate-packages/server-package.zip