shashankprasanna

data = cudf.read_csv(data_dir+'HIGGS.csv', names=col_names, dtype=dtypes_ls)
X_train, X_test, y_train, y_test = train_test_split(data, 'label', train_size=0.70)
  
 hyperparams={
    'n_estimators'       : args.n_estimators,
    'max_depth'          : args.max_depth,
    'n_bins'             : args.n_bins,
    'split_criterion'    : args.split_criterion,
    'split_algo'         : args.split_algo,
    'bootstrap'          : args.bootstrap,

shashankprasanna

parser = argparse.ArgumentParser()

# Hyper-parameters
parser.add_argument('--n_estimators',        type=int,   default=20)
parser.add_argument('--max_depth',           type=int,   default=16)
parser.add_argument('--n_bins',              type=int,   default=8)
parser.add_argument('--split_criterion',     type=int,   default=0)
parser.add_argument('--split_algo',          type=int,   default=0)
parser.add_argument('--bootstrap',           type=bool,  default=True)
parser.add_argument('--bootstrap_features',  type=bool,  default=False)

shashankprasanna

hyperparams={ 
    'n_estimators'       : 15,
    'max_depth'          : 5,
    'n_bins'             : 8,
    'split_criterion'    : 0,      # GINI:0, ENTROPY:1
    'split_algo'         : 0,      # HIST:0 GLOBAL_QUANTILE:1
    'bootstrap'          : 0,      # true: sample with replacement, false: sample without replacement
    'bootstrap_features' : 0,      # true: sample with replacement, false: sample without replacement
    'max_leaves'         : -1,     # unlimited leaves
    'max_features'       : 0.2, 

shashankprasanna

from sagemaker.estimator import Estimator

train_instance_type = 'local_gpu'
local_data_dir = 'file://./dataset'

rapids_estimator = Estimator(image_name='sagemaker-rapids:latest',
                          role=role,
                          train_instance_count=1,
                          train_instance_type=train_instance_type,
                          hyperparameters=hyperparams,

shashankprasanna

from sagemaker.tuner import IntegerParameter, CategoricalParameter, ContinuousParameter, HyperparameterTuner

hyperparameter_ranges = {
    'n_estimators'        : IntegerParameter(10, 200), 
    'max_depth'           : IntegerParameter(1, 22),
    'n_bins'              : IntegerParameter(5, 24),
    'split_criterion'     : CategoricalParameter([0, 1]),
    'split_algo'          : CategoricalParameter([0, 1]),
    'bootstrap'           : CategoricalParameter([True, False]),
    'bootstrap_features'  : CategoricalParameter([True, False]),

shashankprasanna

from sagemaker.estimator import Estimator

train_instance_type = 'ml.p3.2xlarge'
rapids_estimator = Estimator(image_name=image,
                          role=role,
                          train_instance_count=1,
                          train_instance_type=train_instance_type,
                          hyperparameters=hyperparams,
                          metric_definitions=[{'Name': 'test_acc', 'Regex': 'test_acc: ([0-9\\.]+)'}])

shashankprasanna

tuner = HyperparameterTuner(rapids_estimator,
                            objective_metric_name='test_acc',
                            hyperparameter_ranges=hyperparameter_ranges,
                            strategy='Bayesian',
                            max_jobs=1,
                            max_parallel_jobs=1,
                            objective_type='Maximize',
                            metric_definitions=[{'Name': 'test_acc', 'Regex': 'test_acc: ([0-9\\.]+)'}])

shashankprasanna

s3_data_dir = sagemaker_session.upload_data(path='dataset', key_prefix='dataset/higgs-dataset')
job_name = 'rapidsHPO' + time.strftime('%Y-%m-%d-%H-%M-%S-%j', time.gmtime())

tuner.fit({'dataset': s3_data_dir}, job_name=job_name)

shashankprasanna

callbacks = []
callbacks.append(hvd.callbacks.BroadcastGlobalVariablesCallback(0))
callbacks.append(hvd.callbacks.MetricAverageCallback())
callbacks.append(hvd.callbacks.LearningRateWarmupCallback(warmup_epochs=5, verbose=1))
callbacks.append(tf.keras.callbacks.ReduceLROnPlateau(patience=10, verbose=1))
if hvd.rank() == 0:
    callbacks.append(ModelCheckpoint(args.output_data_dir + '/checkpoint-{epoch}.h5'))
    logdir = args.output_data_dir + '/' + datetime.now().strftime("%Y%m%d-%H%M%S")
    callbacks.append(TensorBoard(log_dir=logdir))
    callbacks.append(Sync2S3(logdir=logdir, s3logdir=tensorboard_logs))

shashankprasanna

history = model.fit(train_dataset,
                    steps_per_epoch  = (NUM_TRAIN_IMAGES // batch_size) // size,
                    validation_data  = val_dataset,
                    validation_steps = (NUM_VALID_IMAGES // batch_size) // size,
                    verbose          = 1 if hvd.rank() == 0 else 0,
                    epochs           = epochs, callbacks=callbacks)