Use high-level APIs of SageMaker SDK for PIPE mode augmented manifest training on object_detection_tutorial.ipynb

object_detection_tutorial_high_level_api.py

# object_detection_tutorial.ipynb 

od_model = sagemaker.estimator.Estimator(training_image,
                                         role, 
                                         train_instance_count=1, 
                                         train_instance_type='ml.p3.2xlarge',
                                         train_volume_size = 50,
                                         train_max_run = 360000,
                                         input_mode = 'Pipe',
                                         output_path=s3_output_path, 
                                         sagemaker_session=sess)

od_model.set_hyperparameters(base_network='resnet-50',
                             use_pretrained_model=1,
                             num_classes=80,
                             mini_batch_size=16,
                             epochs=30,
                             learning_rate=0.001,
                             lr_scheduler_step='10',
                             lr_scheduler_factor=0.1,
                             optimizer='sgd',
                             momentum=0.9,
                             weight_decay=0.0005,
                             overlap_threshold=0.5,
                             nms_threshold=0.45,
                             image_shape=512,
                             label_width=600,
                             num_training_samples=4452)

# Create a train data channel with S3_data_type as 'AugmentedManifestFile' and attribute names.
train_data = sagemaker.session.s3_input(s3_train_data_path,
                                        distribution='FullyReplicated',
                                        content_type='application/x-recordio',
                                        record_wrapping='RecordIO', 
                                        s3_data_type='AugmentedManifestFile',
                                        attribute_names=attribute_names) 

validation_data = sagemaker.session.s3_input(s3_validation_data_path,
                                        distribution='FullyReplicated',
                                        content_type='application/x-recordio',
                                        record_wrapping='RecordIO', 
                                        s3_data_type='AugmentedManifestFile',
                                        attribute_names=attribute_names) 

data_channels = {'train': train_data, 'validation': validation_data}

od_model.fit(inputs=data_channels, logs=True)