Kick off Sagemaker Training

sagemaker_train.py

# Notebook Instance Imports
import os
import sagemaker
from sagemaker.tensorflow import TensorFlow

# S3 directories
data_s3 = 's3://jigsaw-toxic-mjy/'
# define inputs
inputs = {'data':data_s3}

# create estimator
estimator = TensorFlow(entry_point='jigsaw_train1_aws2.py', # your training script
                       train_instance_type='ml.p2.xlarge', # instance used for training, usually a GPU instance
                       output_path="s3://jigsaw-toxic-mjy-output", # s3 location to output files
                       train_instance_count=1, # number of instances
                       role=sagemaker.get_execution_role(), # Passes to the container the AWS role that you are using on this notebook
                       framework_version='1.11.0', # Uses TensorFlow 1.11
                       py_version='py3',
                       script_mode=True)

# Run training job
estimator.fit(inputs) # run training