Heavy Hitter

This example demonstrates how to complete a heavy hitter analytics task using local data from multiple clients. The source code locates at [link]. We implemented the Triehh method in the follownig paper:

Federated Heavy Hitters Discovery with Differential Privacy: [paper](https://arxiv.org/pdf/1902.08534.pdf)
Reference code: [code](https://github.com/google-research/federated/tree/master/triehh)

Same as the setting in the paper, we utilize Twitter Sentiment140 dataset.

If you have multiple nodes, you should run the client script on each node

APIs

The highly encapsulated server and client API calls are shown as below:

run_server.sh is as follows:

#!/usr/bin/env bash

RUN_ID=$1
python3 server.py --cf fedml_config.yaml --rank 0 --role server --run_id $RUN_ID

server.py is as follows:

from fedml.fa import init, FARunner
from fedml.fa.data import fa_load_data


if __name__ == "__main__":
    args = init()
    dataset = fa_load_data(args)
    fa_runner = FARunner(args, dataset)
    fa_runner.run()

run_client.sh is as follows:

#!/usr/bin/env bash
RANK=$1
RUN_ID=$2
python3 client.py --cf fedml_config.yaml --rank $RANK --role client --run_id $RUN_ID

client.py is as follows:

from fedml.fa import init, FARunner
from fedml.fa.data import fa_load_data

if __name__ == "__main__":
    args = init()
    dataset = fa_load_data(args)
    fa_runner = FARunner(args, dataset)
    fa_runner.run()

At the client side, the client ID (a.k.a rank) starts from 1. Please also modify fedml_config.yaml, changing the client_num_in_total the as the number of clients you plan to run.

At the server side, run the following script:

bash run_server.sh a

For client 1, run the following script:

bash run_client.sh 1 a

For client 2, run the following script:

bash run_client.sh 2 a

fedml_config.yaml is shown below.

common_args:
  training_type: "cross_silo"
  random_seed: 0

data_args:
  dataset: "twitter"
  data_cache_dir: ~/fedml_data

train_args:
  client_num_in_total: 2
  client_num_per_round: 2
  comm_round: 10
  fa_task: "heavy_hitter"

comm_args:
  backend: "MQTT_S3"

Analyzing Results

At the end of the analysis, the server window will display the following log:

[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:49] [INFO] [__init__.py:19:event] FedMLDebug edge_id = None, event_name = server.wait, START = True
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:49] [INFO] [fedml_server_manager.py:154:handle_message_receive_model_from_client] =============analyzing is finished. Cleanup...============
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:49] [INFO] [mqtt_s3_multi_clients_comm_manager.py:260:send_message] mqtt_s3.send_message: msg topic = fedml_a_0_1
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:49] [INFO] [fedml_server_manager.py:198:send_message_finish] finish from send id 0 to receive id 1.
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:49] [INFO] [fedml_server_manager.py:200:send_message_finish]  ====================send cleanup message to 0====================
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:49] [INFO] [mqtt_s3_multi_clients_comm_manager.py:260:send_message] mqtt_s3.send_message: msg topic = fedml_a_0_2
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:49] [INFO] [fedml_server_manager.py:198:send_message_finish] finish from send id 0 to receive id 2.
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:49] [INFO] [fedml_server_manager.py:200:send_message_finish]  ====================send cleanup message to 1====================
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:52] [INFO] [fedml_comm_manager.py:67:finish] __finish
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:52] [INFO] [mqtt_s3_multi_clients_comm_manager.py:332:stop_receive_message] mqtt_s3.stop_receive_message: stopping...
[FedML-Server @device-id-0] [Wed, 01 Nov 2023 06:46:52] [INFO] [fedml_comm_manager.py:29:run] finished...

At the end of the analysis, the client 1 window will display the following log:

[FedML-Client @device-id-1] [Wed, 01 Nov 2023 06:46:52] [INFO] [fedml_client_master_manager.py:85:handle_message_finish]  ====================cleanup ====================
[FedML-Client @device-id-1] [Wed, 01 Nov 2023 06:46:52] [INFO] [fedml_comm_manager.py:67:finish] __finish
[FedML-Client @device-id-1] [Wed, 01 Nov 2023 06:46:52] [INFO] [mqtt_s3_multi_clients_comm_manager.py:332:stop_receive_message] mqtt_s3.stop_receive_message: stopping...
[FedML-Client @device-id-1] [Wed, 01 Nov 2023 06:46:52] [INFO] [fedml_comm_manager.py:29:run] finished...

At the end of the analysis, the client 2 window will display the following log:

[FedML-Client @device-id-2] [Wed, 01 Nov 2023 06:46:52] [INFO] [fedml_client_master_manager.py:85:handle_message_finish]  ====================cleanup ====================
[FedML-Client @device-id-2] [Wed, 01 Nov 2023 06:46:52] [INFO] [fedml_comm_manager.py:67:finish] __finish
[FedML-Client @device-id-2] [Wed, 01 Nov 2023 06:46:52] [INFO] [mqtt_s3_multi_clients_comm_manager.py:332:stop_receive_message] mqtt_s3.stop_receive_message: stopping...
[FedML-Client @device-id-2] [Wed, 01 Nov 2023 06:46:52] [INFO] [fedml_comm_manager.py:29:run] finished...

Heavy Hitter

APIs​

Analyzing Results​

APIs

Analyzing Results