Skip to content

Latest commit

 

History

History
461 lines (386 loc) · 23.5 KB

quickstart.md

File metadata and controls

461 lines (386 loc) · 23.5 KB
Raw

Quickstart

Note: This guide will not work for Mac M1/M2 systems as they do not support Keras with TensorFlow v1. Refer to the setup.md for the PyTorch guide which will work on Mac M1/M2 systems.

Try out the step-by-step Keras classifier example

All commands are assumed to be run from the base directory at the top of this repository. We also assume other folders, like examples, are located under the same directory.

In this example, we will train a Keras (TensorFlow v1) CNN model, as shown below, on MNIST data in the federated learning fashion.

num_classes = 10
img_rows, img_cols = 28, 28
if K.image_data_format() == 'channels_first':
    input_shape = (1, img_rows, img_cols)
else:
    input_shape = (img_rows, img_cols, 1)

model = Sequential()
model.add(Conv2D(32, (3, 3), activation='relu', input_shape=input_shape))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))

model.compile(loss=keras.losses.categorical_crossentropy,
              optimizer=keras.optimizers.Adadelta(),
              metrics=['accuracy'])

1. Set up a running environment for IBM federated learning

We highly recommend using Conda installation for IBM federated learning. If you don't have Conda, you can install it here.

Once Conda is installed, create a new environment for IBM federated learning by running the following. To use Keras with TensorFlow v1, you must use Python 3.7.

conda create -n <env_name> python=3.7

Once the conda environment is created, activate it by running the following.

conda activate <env_name>

Then install the IBM federated learning package using the wheel file (located in the federated-learning-lib directory) with a Keras backend by running the following.

pip install "/path/to/federated_learning_lib.whl[keras]"

Notes:

  • The quotes are required if using the Zsh shell (this is the default shell for Mac).
  • The latest IBM FL library supports model training using Scikit-learn, PyTorch, Keras (with TensorFlow v1), and TensorFlow v2. You must specify the desired model training backend when installing the IBM FL library. In this example, we specify to install with the Keras backend. See here for more details.

2. Prepare datasets for each participating parties

For example, run

python examples/generate_data.py -n 2 -d mnist -pp 200

This command would generate 2 parties with 200 data points each, randomly sampled from the MNIST dataset. By default, the data is stored under the examples/data/mnist/random directory.

Warning: test set and train set contain different labels
Party_ 0
nb_x_train:  (200, 28, 28) nb_x_test:  (5000, 28, 28)
* Label  0  samples:  22
* Label  1  samples:  30
* Label  2  samples:  16
* Label  3  samples:  22
* Label  4  samples:  17
* Label  5  samples:  25
* Label  6  samples:  12
* Label  7  samples:  15
* Label  8  samples:  19
* Label  9  samples:  22
Finished! :) Data saved in  examples/data/mnist/random
Party_ 1
nb_x_train:  (200, 28, 28) nb_x_test:  (5000, 28, 28)
* Label  0  samples:  22
* Label  1  samples:  18
* Label  2  samples:  22
* Label  3  samples:  23
* Label  4  samples:  14
* Label  5  samples:  23
* Label  6  samples:  22
* Label  7  samples:  20
* Label  8  samples:  18
* Label  9  samples:  18
Finished! :) Data saved in  examples/data/mnist/random

For a full description of the different options to prepare datasets, run python examples/generate_data.py -h.

3. Define model specification and create configuration files for the aggregator and parties

For example, run:

python examples/generate_configs.py -n <num_parties> -f iter_avg -m keras -d mnist -p <path>

This command performs two tasks:

  1. It specifies the machine learning model to be trained, in this case, a Keras CNN classifier.

  2. It generates the configuration files necessary to train a keras model via fusion algorithm iter_avg, assuming <num_parties> parties join the federated learning training. You must also specify the dataset name via -d and the party data path via -p.

In this example, we run:

python examples/generate_configs.py -n 2 -f iter_avg -m keras -d mnist -p examples/data/mnist/random/

Hence, we generate 2 parties in our example, using the mnist dataset and examples/data/mnist/random as our data path.

Finished generating config file for aggregator. Files can be found in:  <whl_directory>/examples/configs/iter_avg/keras/config_agg.yml
Finished generating config file for parties. Files can be found in:  <whl_directory>/examples/configs/iter_avg/keras/config_party*.yml

You may also see warning messages which are fine. For a full description of the different options, run python examples/generate_configs.py -h.

Below you can see samples of configuration files.

  • Aggregator's configuration file:
connection:
  info:
    ip: 127.0.0.1
    port: 5000
    tls_config:
      enable: false
  name: FlaskConnection
  path: ibmfl.connection.flask_connection
  sync: false
data:
  info:
    npz_file: examples/datasets/mnist.npz
  name: MnistKerasDataHandler
  path: ibmfl.util.data_handlers.mnist_keras_data_handler
fusion:
  name: IterAvgFusionHandler
  path: ibmfl.aggregator.fusion.iter_avg_fusion_handler
hyperparams:
  global:
    max_timeout: 60
    parties: 2
    rounds: 3
    termination_accuracy: 0.9
  local:
    optimizer:
      lr: 0.01
    training:
      epochs: 3
protocol_handler:
  name: ProtoHandler
  path: ibmfl.aggregator.protohandler.proto_handler
  • Party's configuration file:
aggregator:
  ip: 127.0.0.1
  port: 5000
connection:
  info:
    ip: 127.0.0.1
    port: 8085
    tls_config:
      enable: false
  name: FlaskConnection
  path: ibmfl.connection.flask_connection
  sync: false
data:
  info:
    npz_file: examples/data/mnist/random/data_party0.npz
  name: MnistKerasDataHandler
  path: ibmfl.util.data_handlers.mnist_keras_data_handler
local_training:
  name: LocalTrainingHandler
  path: ibmfl.party.training.local_training_handler
model:
  name: KerasFLModel
  path: ibmfl.model.keras_fl_model
  spec:
    model_definition: examples/configs/iter_avg/keras/compiled_keras.h5
    model_name: keras-cnn
protocol_handler:
  name: PartyProtocolHandler
  path: ibmfl.party.party_protocol_handler

Notice that the configuration files contain a data section that is different for each party. In fact, each party's points to its own data, generated from the command in step 2.

4. Start the aggregator

To start the aggregator, open a terminal window running the IBM federated fearning environment set up beforehand, and check that you are in the correct directory. In the terminal run:

python -m ibmfl.aggregator.aggregator examples/configs/iter_avg/keras/config_agg.yml

where the path provided is the aggregator configuration file path.

2020-06-29 11:38:38,058 - ibmfl.util.config - INFO - Getting details from config file.
2020-06-29 11:38:45,001 - ibmfl.util.config - INFO - No model config provided for this setup.
2020-06-29 11:38:45,353 - ibmfl.util.config - INFO - No local training config provided for this setup.
2020-06-29 11:38:45,353 - ibmfl.connection.flask_connection - INFO - RestSender initialized
2020-06-29 11:38:45,353 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.START
<ibmfl.connection.router_handler.Router object at 0x106cb9630>
2020-06-29 11:38:45,353 - ibmfl.connection.flask_connection - INFO - Receiver Initialized
2020-06-29 11:38:45,353 - ibmfl.connection.flask_connection - INFO - Initializing Flask application
2020-06-29 11:38:45,356 - __main__ - INFO - Aggregator initialization successful

Then in the terminal, type START and press enter.

START
2020-06-29 11:39:24,349 - root - INFO - State: States.CLI_WAIT
2020-06-29 11:39:24,349 - __main__ - INFO - Aggregator start successful
 * Serving Flask app "ibmfl.connection.flask_connection" (lazy loading)
 * Environment: production
   WARNING: This is a development server. Do not use it in a production deployment.
   Use a production WSGI server instead.
 * Debug mode: off
2020-06-29 11:39:24,356 - werkzeug - INFO -  * Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)

5. Start and register parties

To start and register a new party, open one new terminal window for each party, running the IBM federated learning environment set up beforehand, and make sure you are in the correct directory. In the terminal run:

python -m ibmfl.party.party examples/configs/iter_avg/keras/config_party<idx>.yml

where the path provided is the path to the party's configuration file.

Note: Each party will have a different configuration file; in our example, it is noted by changing config_party<idx>.yml. For instance, to start the 1st party, one would run:

python -m ibmfl.party.party examples/configs/iter_avg/keras/config_party0.yml
2020-06-29 11:40:30,420 - ibmfl.util.config - INFO - Getting config from file
2020-06-29 11:40:30,420 - ibmfl.util.config - INFO - Getting details from config file.
2020-06-29 11:40:33,617 - ibmfl.util.config - INFO - No fusion config provided for this setup.
2020-06-29 11:40:33,987 - ibmfl.connection.flask_connection - INFO - RestSender initialized
<ibmfl.connection.router_handler.Router object at 0x135de64a8>
2020-06-29 11:40:33,997 - ibmfl.connection.flask_connection - INFO - Receiver Initialized
2020-06-29 11:40:33,997 - ibmfl.connection.flask_connection - INFO - Initializing Flask application
2020-06-29 11:40:34,000 - __main__ - INFO - Party initialization successful

You may also see warning messages which are fine. In the terminal for each party, type START and press enter to start the party. Then type REGISTER and press enter to register the party for the federated learning task.

START
2020-06-29 11:41:35,732 - __main__ - INFO - Party start successful
 * Serving Flask app "ibmfl.connection.flask_connection" (lazy loading)
 * Environment: production
   WARNING: This is a development server. Do not use it in a production deployment.
   Use a production WSGI server instead.
 * Debug mode: off
2020-06-29 11:41:35,735 - werkzeug - INFO -  * Running on http://127.0.0.1:8085/ (Press CTRL+C to quit)
REGISTER
2020-06-29 11:41:47,056 - __main__ - INFO - Registering party...
2020-06-29 11:41:47,071 - __main__ - INFO - Registration Successful

The aggregator terminal will also prompt out INFO to show that it receives the party's registration message (as shown in the third figure on the right).

2020-06-29 11:41:47,067 - ibmfl.connection.flask_connection - INFO - Request received for path :6
2020-06-29 11:41:47,068 - ibmfl.aggregator.protohandler.proto_handler - INFO - Adding party with id 78da36fb-2443-45d0-80d2-b438bf8c01a5
2020-06-29 11:41:47,069 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:41:47] "POST /6 HTTP/1.1" 200 -

6. Initiate training from the aggregator

To initiate federated training, type TRAIN in your aggregator terminal and press enter.

Note: In this example, we have 2 parties join the training and we run 3 global rounds, each round with 3 local epochs.

Outputs in the aggregator terminal after running the above command will look like:

TRAIN
2020-06-29 11:43:43,982 - root - INFO - State: States.PROC_TRAIN
2020-06-29 11:43:43,982 - __main__ - INFO - Initiating Global Training.
2020-06-29 11:43:43,982 - ibmfl.aggregator.fusion.fusion_handler - INFO - Warm start disabled.
2020-06-29 11:43:43,982 - ibmfl.aggregator.fusion.iter_avg_fusion_handler - INFO - Model updateNone
2020-06-29 11:43:43,983 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.SND_REQ
2020-06-29 11:43:44,090 - ibmfl.aggregator.protohandler.proto_handler - INFO - Total number of success responses :2
2020-06-29 11:43:44,090 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.QUORUM_WAIT
2020-06-29 11:43:44,090 - ibmfl.aggregator.protohandler.proto_handler - INFO - Target Qorum: 2
2020-06-29 11:43:47,423 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:47,429 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:47,480 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:47] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:47,518 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:47] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:49,094 - ibmfl.aggregator.protohandler.proto_handler - INFO - Timeout:60 Time spent:5
2020-06-29 11:43:49,094 - ibmfl.aggregator.protohandler.proto_handler - INFO - Target Qorum: 2
2020-06-29 11:43:49,094 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.PROC_RSP
2020-06-29 11:43:49,108 - ibmfl.aggregator.fusion.iter_avg_fusion_handler - INFO - Model update<ibmfl.model.model_update.ModelUpdate object at 0x111835828>
2020-06-29 11:43:49,108 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.SND_REQ
2020-06-29 11:43:49,318 - ibmfl.aggregator.protohandler.proto_handler - INFO - Total number of success responses :2
2020-06-29 11:43:49,318 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.QUORUM_WAIT
2020-06-29 11:43:49,318 - ibmfl.aggregator.protohandler.proto_handler - INFO - Target Qorum: 2
2020-06-29 11:43:51,096 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:51,131 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:51] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:51,133 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:51,176 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:51] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:54,321 - ibmfl.aggregator.protohandler.proto_handler - INFO - Timeout:60 Time spent:5
2020-06-29 11:43:54,321 - ibmfl.aggregator.protohandler.proto_handler - INFO - Target Qorum: 2
2020-06-29 11:43:54,321 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.PROC_RSP
2020-06-29 11:43:54,341 - ibmfl.aggregator.fusion.iter_avg_fusion_handler - INFO - Model update<ibmfl.model.model_update.ModelUpdate object at 0x120ceb128>
2020-06-29 11:43:54,341 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.SND_REQ
2020-06-29 11:43:54,550 - ibmfl.aggregator.protohandler.proto_handler - INFO - Total number of success responses :2
2020-06-29 11:43:54,550 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.QUORUM_WAIT
2020-06-29 11:43:54,550 - ibmfl.aggregator.protohandler.proto_handler - INFO - Target Qorum: 2
2020-06-29 11:43:56,553 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:56,583 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:56] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:56,584 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:56,615 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:56] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:59,553 - ibmfl.aggregator.protohandler.proto_handler - INFO - Timeout:60 Time spent:5
2020-06-29 11:43:59,553 - ibmfl.aggregator.protohandler.proto_handler - INFO - Target Qorum: 2
2020-06-29 11:43:59,553 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.PROC_RSP
2020-06-29 11:43:59,571 - ibmfl.aggregator.fusion.iter_avg_fusion_handler - INFO - Reached maximum global rounds. Finish training :) 
2020-06-29 11:43:59,572 - __main__ - INFO - Finished Global Training

Outputs in party's (party 1) terminal after running the above command will look like:

2020-06-29 11:43:43,998 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:43,999 - ibmfl.party.party_protocol_handler - INFO - received a async request
2020-06-29 11:43:43,999 - ibmfl.party.party_protocol_handler - INFO - finished async request
2020-06-29 11:43:44,000 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:44] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:44,000 - ibmfl.party.party_protocol_handler - INFO - Handling async request in a separate thread
2020-06-29 11:43:44,001 - ibmfl.party.party_protocol_handler - INFO - Received request from aggregator
2020-06-29 11:43:44,001 - ibmfl.party.party_protocol_handler - INFO - Received request in with message_type:  7
2020-06-29 11:43:44,001 - ibmfl.party.party_protocol_handler - INFO - Received request in PH 7
2020-06-29 11:43:44,001 - ibmfl.util.data_handlers.mnist_keras_data_handler - INFO - Loaded training data from examples/data/mnist/random/data_party0.npz
x_train shape: (200, 28, 28, 1)
200 train samples
5000 test samples
2020-06-29 11:43:44,052 - ibmfl.party.training.local_training_handler - INFO - No model update was provided.
2020-06-29 11:43:44,053 - ibmfl.party.training.local_training_handler - INFO - Local training started...
2020-06-29 11:43:44,053 - ibmfl.model.keras_fl_model - INFO - Using default hyperparameters:  batch_size:128
2020-06-29 11:43:44,080 - tensorflow - WARNING - From /Users/[email protected]/opt/anaconda3/envs/ibmfl/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.cast instead.
Epoch 1/3
200/200 [==============================] - 1s 5ms/step - loss: 2.2764 - acc: 0.1150
Epoch 2/3
200/200 [==============================] - 1s 4ms/step - loss: 2.0754 - acc: 0.2400
Epoch 3/3
200/200 [==============================] - 1s 6ms/step - loss: 1.7741 - acc: 0.4000
2020-06-29 11:43:47,353 - ibmfl.party.training.local_training_handler - INFO - Local training done, generating model update...
2020-06-29 11:43:47,359 - ibmfl.party.party_protocol_handler - INFO - successfully finished async request
2020-06-29 11:43:49,200 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:49,236 - ibmfl.party.party_protocol_handler - INFO - received a async request
2020-06-29 11:43:49,236 - ibmfl.party.party_protocol_handler - INFO - finished async request
2020-06-29 11:43:49,237 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:49] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:49,238 - ibmfl.party.party_protocol_handler - INFO - Handling async request in a separate thread
2020-06-29 11:43:49,238 - ibmfl.party.party_protocol_handler - INFO - Received request from aggregator
2020-06-29 11:43:49,238 - ibmfl.party.party_protocol_handler - INFO - Received request in with message_type:  7
2020-06-29 11:43:49,238 - ibmfl.party.party_protocol_handler - INFO - Received request in PH 7
2020-06-29 11:43:49,238 - ibmfl.util.data_handlers.mnist_keras_data_handler - INFO - Loaded training data from examples/data/mnist/random/data_party0.npz
x_train shape: (200, 28, 28, 1)
200 train samples
5000 test samples
2020-06-29 11:43:49,277 - ibmfl.party.training.local_training_handler - INFO - Local model updated.
2020-06-29 11:43:49,277 - ibmfl.party.training.local_training_handler - INFO - Local training started...
2020-06-29 11:43:49,277 - ibmfl.model.keras_fl_model - INFO - Using default hyperparameters:  batch_size:128
Epoch 1/3
200/200 [==============================] - 1s 4ms/step - loss: 1.4924 - acc: 0.6300
Epoch 2/3
200/200 [==============================] - 1s 3ms/step - loss: 1.8330 - acc: 0.4100
Epoch 3/3
200/200 [==============================] - 0s 2ms/step - loss: 1.0651 - acc: 0.6800
2020-06-29 11:43:51,074 - ibmfl.party.training.local_training_handler - INFO - Local training done, generating model update...
2020-06-29 11:43:51,078 - ibmfl.party.party_protocol_handler - INFO - successfully finished async request
2020-06-29 11:43:54,418 - ibmfl.connection.flask_connection - INFO - Request received for path :7
2020-06-29 11:43:54,455 - ibmfl.party.party_protocol_handler - INFO - received a async request
2020-06-29 11:43:54,456 - ibmfl.party.party_protocol_handler - INFO - finished async request
2020-06-29 11:43:54,458 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:43:54] "POST /7 HTTP/1.1" 200 -
2020-06-29 11:43:54,459 - ibmfl.party.party_protocol_handler - INFO - Handling async request in a separate thread
2020-06-29 11:43:54,459 - ibmfl.party.party_protocol_handler - INFO - Received request from aggregator
2020-06-29 11:43:54,459 - ibmfl.party.party_protocol_handler - INFO - Received request in with message_type:  7
2020-06-29 11:43:54,459 - ibmfl.party.party_protocol_handler - INFO - Received request in PH 7
2020-06-29 11:43:54,459 - ibmfl.util.data_handlers.mnist_keras_data_handler - INFO - Loaded training data from examples/data/mnist/random/data_party0.npz
x_train shape: (200, 28, 28, 1)
200 train samples
5000 test samples
2020-06-29 11:43:54,502 - ibmfl.party.training.local_training_handler - INFO - Local model updated.
2020-06-29 11:43:54,502 - ibmfl.party.training.local_training_handler - INFO - Local training started...
2020-06-29 11:43:54,502 - ibmfl.model.keras_fl_model - INFO - Using default hyperparameters:  batch_size:128
Epoch 1/3
200/200 [==============================] - 0s 2ms/step - loss: 0.9385 - acc: 0.7250
Epoch 2/3
200/200 [==============================] - 1s 4ms/step - loss: 1.0080 - acc: 0.7200
Epoch 3/3
200/200 [==============================] - 1s 4ms/step - loss: 0.7465 - acc: 0.7950
2020-06-29 11:43:56,514 - ibmfl.party.training.local_training_handler - INFO - Local training done, generating model update...
2020-06-29 11:43:56,518 - ibmfl.party.party_protocol_handler - INFO - successfully finished async request

Outputs from party 2 will be similar as party 1.

7. (Optional) Issue various commands to train again, evaluate, sync, and save the models

For a full list of supported commands, see examples/README.md. Sample outputs of issuing the EVAL command in one of the parties' terminal after the global training.

EVAL
2020-06-29 11:46:05,003 - ibmfl.util.data_handlers.mnist_keras_data_handler - INFO - Loaded training data from examples/data/mnist/random/data_party0.npz
x_train shape: (200, 28, 28, 1)
200 train samples
5000 test samples
5000/5000 [==============================] - 2s 303us/step
2020-06-29 11:46:06,542 - ibmfl.party.party_protocol_handler - INFO - {'loss': 0.6104391970634461, 'acc': 0.8152}

Users can also enter TRAIN again at the aggregator's terminal if they want to continue the FL training. Entering SYNC at the aggregator's terminal will trigger the synchronization of the current global model with parties, and SAVE will trigger the parties to save their models at the local working directory.

8. Terminate the aggregator and parties processes

Remember to use STOP to terminate the aggregator's and parties' processes and exit. Outputs in the aggregator terminal after running STOP:

STOP
2020-06-29 11:46:44,476 - root - INFO - State: States.PROC_STOP
2020-06-29 11:46:44,476 - ibmfl.aggregator.protohandler.proto_handler - INFO - State: States.SND_REQ
2020-06-29 11:46:44,581 - ibmfl.aggregator.protohandler.proto_handler - INFO - Total number of success responses :2
2020-06-29 11:46:44,582 - ibmfl.connection.flask_connection - INFO - Stopping Receiver and Sender
2020-06-29 11:46:44,584 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:46:44] "POST /shutdown HTTP/1.1" 200 -
2020-06-29 11:46:44,585 - __main__ - INFO - Aggregator stop successful

Outputs in the party's terminal after running STOP:

STOP
2020-06-29 11:47:01,587 - ibmfl.connection.flask_connection - INFO - Stopping Receiver and Sender
2020-06-29 11:47:01,591 - werkzeug - INFO - 127.0.0.1 - - [29/Jun/2020 11:47:01] "POST /shutdown HTTP/1.1" 200 -