Python TorchFederatedLearnerTechnicalConfig Examples

Example #1

File: run_10_clients.py Project: matech96/FederatedLearning

def do_training(config: TorchFederatedLearnerCIFAR100Config):
    config_technical = TorchFederatedLearnerTechnicalConfig(
        STORE_OPT_ON_DISK=False,
        STORE_MODEL_IN_RAM=False,
    )

    name = f"{config.SERVER_OPT}: {config.SERVER_LEARNING_RATE} - {config.CLIENT_OPT_STRATEGY} - {config.CLIENT_OPT}: {config.CLIENT_LEARNING_RATE}"
    logging.info(name)
    experiment = Experiment(workspace="federated-learning",
                            project_name=project_name)
    experiment.set_name(name)
    learner = TorchFederatedLearnerCIFAR100(experiment, config,
                                            config_technical)
    learner.train()

Example #2

            CLIENT_OPT=common.get_name(client_opt),
            CLIENT_OPT_ARGS=common.get_args(client_opt),
            # CLIENT_OPT_L2=1e-4,
            CLIENT_OPT_STRATEGY=client_opt_strategy,
            SERVER_OPT=common.get_name(server_opt),
            SERVER_OPT_ARGS=common.get_args(server_opt),
            SERVER_LEARNING_RATE=server_lr,
            IS_IID_DATA=is_iid,
            BATCH_SIZE=B,
            CLIENT_FRACTION=C,
            N_CLIENTS=NC,
            N_EPOCH_PER_CLIENT=E,
            MAX_ROUNDS=max_rounds,
            MODEL=model,
            SCAFFOLD=True)
        config_technical = TorchFederatedLearnerTechnicalConfig(
            BREAK_ROUND=300,
            EVAL_ROUND=1,
            TEST_LAST=1,
            # STORE_OPT_ON_DISK=False,
            # STORE_MODEL_IN_RAM=False,
        )
        name = f"{config.SERVER_OPT}: {config.SERVER_LEARNING_RATE} - {config.CLIENT_OPT_STRATEGY} - {config.CLIENT_OPT}: {config.CLIENT_LEARNING_RATE}"
        experiment = Experiment(workspace="federated-learning-emnist-m",
                                project_name=project_name)
        try:
            common.do_training_emnist(experiment, name, config,
                                      config_technical)
        except ToLargeLearningRateExcpetion:
            pass

Example #3

File: run_baseline.py Project: matech96/FederatedLearning

client_opt = "SGD"
client_opt_strategy = "reinit"
# image_norm = "tflike"
# TODO a paraméterek helytelen nevére nem adott hibát
config = TorchFederatedLearnerCIFAR100Config(
    BREAK_ROUND=1500,
    CLIENT_LEARNING_RATE=client_lr,
    CLIENT_OPT=client_opt,
    # CLIENT_OPT_ARGS=common.get_args(client_opt),
    CLIENT_OPT_L2=1e-4,
    CLIENT_OPT_STRATEGY=client_opt_strategy,
    SERVER_OPT=server_opt,
    SERVER_OPT_ARGS=common.get_args(server_opt),
    SERVER_LEARNING_RATE=server_lr,
    IS_IID_DATA=is_iid,
    BATCH_SIZE=B,
    CLIENT_FRACTION=C,
    N_CLIENTS=NC,
    N_EPOCH_PER_CLIENT=E,
    MAX_ROUNDS=max_rounds,
    IMAGE_NORM="recordwisefull",
    NORM="group",
    INIT="tffed",
    AUG="basicf")
config_technical = TorchFederatedLearnerTechnicalConfig(BREAK_ROUND=300,
                                                        EVAL_ROUND=100)
name = f"{config.SERVER_OPT}: {config.SERVER_LEARNING_RATE} - {config.CLIENT_OPT_STRATEGY} - {config.CLIENT_OPT}: {config.CLIENT_LEARNING_RATE}"
experiment = Experiment(workspace="federated-learning",
                        project_name=project_name)
common.do_training(experiment, name, config, config_technical)

Example #4

File: explore_config_10c5e.py Project: matech96/FederatedLearning

server_lr = 0.01
client_lr = 0.01
server_opt = "SGD"
client_opt = "SGD"
client_opt_strategy = "reinit"
project_name = f"10c40e-s-{server_opt}-c-{client_opt}"

max_rounds = 30  # 1500
C = 0.5  # 10 / 500
NC = 10  # 500
E = 40
B = 20
is_iid = False

config_technical = TorchFederatedLearnerTechnicalConfig()

config = TorchFederatedLearnerCIFAR100Config(
    BREAK_ROUND=300,
    CLIENT_LEARNING_RATE=client_lr,
    CLIENT_OPT=client_opt,
    # CLIENT_OPT_ARGS=common.get_args(client_opt),
    CLIENT_OPT_L2=1e-4,
    CLIENT_OPT_STRATEGY=client_opt_strategy,
    SERVER_OPT=server_opt,
    # SERVER_OPT_ARGS=common.get_args(server_opt),
    SERVER_LEARNING_RATE=server_lr,
    IS_IID_DATA=is_iid,
    BATCH_SIZE=B,
    CLIENT_FRACTION=C,
    N_CLIENTS=NC,

Example #5

server_lr = 1.0
client_lr = 0.1
server_opt = "SGD"
client_opt = "SGD"
client_opt_strategy = "reinit"

max_rounds = 30
n_clients_per_round = 10
NC = 10
C = n_clients_per_round / NC
E = 10
B = 20
is_iid = False
project_name = f"{NC}c{E}e{max_rounds}r{n_clients_per_round}f-{server_opt}-{client_opt_strategy[0]}-{client_opt}"

config_technical = TorchFederatedLearnerTechnicalConfig(
    BREAK_ROUND=3, STORE_OPT_ON_DISK=False, STORE_MODEL_IN_RAM=False)

config = TorchFederatedLearnerCIFAR100Config(
    CLIENT_LEARNING_RATE=client_lr,
    CLIENT_OPT=common.get_name(client_opt),
    CLIENT_OPT_ARGS=common.get_args(client_opt),
    CLIENT_OPT_L2=1e-4,
    CLIENT_OPT_STRATEGY=client_opt_strategy,
    SERVER_OPT=common.get_name(server_opt),
    SERVER_OPT_ARGS=common.get_args(server_opt),
    SERVER_LEARNING_RATE=server_lr,
    IS_IID_DATA=is_iid,
    BATCH_SIZE=B,
    CLIENT_FRACTION=C,
    N_CLIENTS=NC,
    N_EPOCH_PER_CLIENT=E,

Example #6

File: run_emnist100s_server.py Project: matech96/FederatedLearning

            CLIENT_OPT=common.get_name(client_opt),
            CLIENT_OPT_ARGS=common.get_args(client_opt),
            # CLIENT_OPT_L2=1e-4,
            CLIENT_OPT_STRATEGY=client_opt_strategy,
            SERVER_OPT=common.get_name(server_opt),
            SERVER_OPT_ARGS=common.get_args(server_opt),
            SERVER_LEARNING_RATE=server_lr,
            IS_IID_DATA=is_iid,
            BATCH_SIZE=B,
            CLIENT_FRACTION=C,
            N_CLIENTS=NC,
            N_EPOCH_PER_CLIENT=E,
            MAX_ROUNDS=max_rounds,
            MODEL=model,
        )
        config_technical = TorchFederatedLearnerTechnicalConfig(
            BREAK_ROUND=300,
            EVAL_ROUND=10,
            TEST_LAST=20,
            STORE_OPT_ON_DISK=True,
            STORE_MODEL_IN_RAM=True,
        )
        name = f"{config.SERVER_OPT}: {config.SERVER_LEARNING_RATE} - {config.CLIENT_OPT_STRATEGY} - {config.CLIENT_OPT}: {config.CLIENT_LEARNING_RATE}"
        experiment = Experiment(workspace="federated-learning-emnist-s",
                                project_name=project_name)
        try:
            common.do_training_emnist(experiment, name, config,
                                      config_technical)
        except ToLargeLearningRateExcpetion:
            pass

Example #7

config_changes = [
    ("SGD", "SGD", 1, 0.1, "nothing"),
    ("Yogi", "SGD", 0.1, 0.01, "nothing"),
    ("Yogi", "Yogi", 0.1, 0.0001, "avg"),
    ("Yogi", "Yogi", 0.1, 0.0001, "reinit"),
    ("Yogi", "Yogi", 0.1, 0.0001, "nothing"),
]
for values in config_changes:
    config = TorchFederatedLearnerCIFAR100Config(
        BREAK_ROUND=1500,
        CLIENT_OPT_L2=1e-4,
        IS_IID_DATA=is_iid,
        BATCH_SIZE=B,
        CLIENT_FRACTION=C,
        N_CLIENTS=NC,
        N_EPOCH_PER_CLIENT=E,
        MAX_ROUNDS=max_rounds,
        IMAGE_NORM="recordwisefull",
        NORM="group",
        INIT="tffed",
        AUG="flipf",
    )
    for k, v in zip(param_names, values):
        setattr(config, k, v)
    config_technical = TorchFederatedLearnerTechnicalConfig(
        SAVE_CHP_INTERVALL=5)
    name = f"{config.SERVER_OPT}: {config.SERVER_LEARNING_RATE} - {config.CLIENT_OPT_STRATEGY} - {config.CLIENT_OPT}: {config.CLIENT_LEARNING_RATE}"
    experiment = Experiment(workspace="federated-learning",
                            project_name=project_name)
    common.do_training(experiment, name, config, config_technical)

Example #8

File: batch_run_explore_10c_e.py Project: matech96/FederatedLearning

    ("SGD", "SGD", 1, 0.1, "reinit"),
    ("Yogi", "SGD", 0.1, 0.01, "reinit"),
    ("Yogi", "Yogi", 0.1, 0.0001, "avg"),
    ("Yogi", "Yogi", 0.1, 0.0001, "reinit"),
    ("Yogi", "Yogi", 0.1, 0.0001, "nothing"),
]
for values in config_changes:
    project_name = f"{NC}c{E}e-{values[0]}-{values[4]}-{values[1]}"
    config = TorchFederatedLearnerCIFAR100Config(
        BREAK_ROUND=5,
        CLIENT_OPT_L2=1e-4,
        IS_IID_DATA=is_iid,
        BATCH_SIZE=B,
        CLIENT_FRACTION=C,
        N_CLIENTS=NC,
        N_EPOCH_PER_CLIENT=E,
        MAX_ROUNDS=max_rounds,
        IMAGE_NORM="recordwisefull",
        NORM="group",
        INIT="tffed",
        AUG="flipf",
    )
    for k, v in zip(param_names, values):
        setattr(config, k, v)
    config_technical = TorchFederatedLearnerTechnicalConfig(
        SAVE_CHP_INTERVALL=5,
        STORE_OPT_ON_DISK=False,
        STORE_MODEL_IN_RAM=False)
    explore_lr(project_name, TorchFederatedLearnerCIFAR100, config,
               config_technical)

Example #9

        CLIENT_LEARNING_RATE=client_lr,
        CLIENT_OPT=common.get_name(client_opt),
        CLIENT_OPT_ARGS=common.get_args(client_opt),
        CLIENT_OPT_L2=1e-4,
        # CLIENT_OPT_STRATEGY=client_opt_strategy,
        SERVER_OPT=common.get_name(server_opt),
        SERVER_OPT_ARGS=common.get_args(server_opt),
        SERVER_LEARNING_RATE=server_lr,
        IS_IID_DATA=is_iid,
        BATCH_SIZE=B,
        CLIENT_FRACTION=C,
        N_CLIENTS=NC,
        N_EPOCH_PER_CLIENT=E,
        MAX_ROUNDS=max_rounds,
        IMAGE_NORM="recordwisefull",
        NORM="group",
        INIT="tffed",
        AUG="basicf",
    )
    if len(param_names) == 1:
        setattr(config, param_names[0], values)
    else:
        for k, v in zip(param_names, values):
            setattr(config, k, v)
    config_technical = TorchFederatedLearnerTechnicalConfig(
        SAVE_CHP_INTERVALL=5, BREAK_ROUND=3)
    name = f"{config.SERVER_OPT}: {config.SERVER_LEARNING_RATE} - {config.CLIENT_OPT_STRATEGY} - {config.CLIENT_OPT}: {config.CLIENT_LEARNING_RATE}"
    experiment = Experiment(workspace="federated-learning-hpopt",
                            project_name=project_name)
    common.do_training(experiment, name, config, config_technical)

Example #10

E = 1
project_name = f"{model}{NC}c{E}e{max_rounds}r{n_clients_per_round}f-{server_opt}-{client_opt_strategy[0]}-{client_opt}"

config = TorchFederatedLearnerEMNISTConfig(
    CLIENT_LEARNING_RATE=client_lr,
    CLIENT_OPT=common.get_name(client_opt),
    CLIENT_OPT_ARGS=common.get_args(client_opt),
    CLIENT_OPT_L2=1e-4,
    CLIENT_OPT_STRATEGY=client_opt_strategy,
    SERVER_OPT=common.get_name(server_opt),
    SERVER_OPT_ARGS=common.get_args(server_opt),
    SERVER_LEARNING_RATE=server_lr,
    IS_IID_DATA=is_iid,
    BATCH_SIZE=B,
    CLIENT_FRACTION=C,
    N_CLIENTS=NC,
    N_EPOCH_PER_CLIENT=E,
    MAX_ROUNDS=max_rounds,
    MODEL=model,
    # SCAFFOLD=True,
)
config_technical = TorchFederatedLearnerTechnicalConfig(
    BREAK_ROUND=300, EVAL_ROUND=10, TEST_LAST=20, STORE_OPT_ON_DISK=False)
name = f"{config.SERVER_OPT}: {config.SERVER_LEARNING_RATE} - {config.CLIENT_OPT_STRATEGY} - {config.CLIENT_OPT}: {config.CLIENT_LEARNING_RATE}"
experiment = Experiment(workspace="federated-learning-scaffold",
                        project_name=project_name)
try:
    common.do_training_emnist(experiment, name, config, config_technical)
except ToLargeLearningRateExcpetion:
    pass

Example #11

File: explore_l2.py Project: matech96/FederatedLearning

for l2 in [1e-3, 1e-2, 1e-1, 1e-1, 1e1]:
    s_opt_args = common.get_args(server_opt)
    s_opt_args["weight_decay"] = l2
    config = TorchFederatedLearnerCIFAR100Config(
        BREAK_ROUND=300,
        CLIENT_LEARNING_RATE=client_lr,
        CLIENT_OPT=client_opt,
        CLIENT_OPT_ARGS=common.get_args(client_opt),
        CLIENT_OPT_L2=l2,
        CLIENT_OPT_STRATEGY=client_opt_strategy,
        SERVER_OPT=server_opt,
        SERVER_OPT_ARGS=s_opt_args,
        SERVER_LEARNING_RATE=server_lr,
        IS_IID_DATA=is_iid,
        BATCH_SIZE=B,
        CLIENT_FRACTION=C,
        N_CLIENTS=NC,
        N_EPOCH_PER_CLIENT=E,
        MAX_ROUNDS=max_rounds,
        DL_N_WORKER=0,
        NORM="group",
        # IMAGE_NORM=image_norm,
        INIT="tffed",
    )
    config_technical = TorchFederatedLearnerTechnicalConfig(HIST_SAMPLE=0)
    name = f"{config.SERVER_OPT}: {config.SERVER_LEARNING_RATE} - {config.CLIENT_OPT_STRATEGY} - {config.CLIENT_OPT}: {config.CLIENT_LEARNING_RATE}"
    experiment = Experiment(workspace="federated-learning",
                            project_name=project_name)
    common.do_training(experiment, name, config, config_technical)