EniasCailliau

from steamship import Steamship

bible = Steamship.use("audio-analytics", "joe-rogan-bible")
bible.analyze_youtube(YOUTUBE_URL)

# Then, later, query like this
bible.query("""
	kind "sentiment" and name "NEGATIVE" 
	overlaps {
		kind "entity" and name "white powder"

EniasCailliau

from steamship import Steamship, File 

workspace = Steamship(workspace="joe-rogan-bible")
with workspace:
  f = File.create("youtube-importer", podcast_url)
  f.transcribe().tag("entities").tag("sentiments").tag("topics").tag("summaries")

EniasCailliau

predictor = estimator.deploy(initial_instance_count=1, instance_type='ml.m4.xlarge')

import numpy as np
random_image_data = np.random.rand(28, 28, 1)
predictor.predict(random_image_data)

EniasCailliau

model_artifacts_location = f's3://{bucket}/{prefix}/artifacts'

estimator = TensorFlow(entry_point='cnn_fashion_mnist.py',
                      role=role,
                      input_mode='Pipe',
                      output_path=model_artifacts_location,
                      training_steps=20000, 
                      evaluation_steps=100,
                      train_instance_count=1,
                      train_instance_type='ml.p2.xlarge',

EniasCailliau

train_data = 's3://sagemaker-eu-central-1-959924085179/radix/mnist_fashion_tutorial/data/mnist/train.tfrecords'
eval_data = 's3://sagemaker-eu-central-1-959924085179/radix/mnist_fashion_tutorial/data/mnist/validation.tfrecords'

tuner.fit({'train': train_data, 'eval': eval_data}, logs=False)

EniasCailliau

estimator = TensorFlow(entry_point='cnn_fashion_mnist.py',
                       role=role,
                       input_mode='Pipe',
                       training_steps=20000, 
                       evaluation_steps=100,
                       train_instance_count=1,
                       train_instance_type='ml.c5.2xlarge',
                       base_job_name='radix_mnist_fashion')

EniasCailliau

# Define objective
objective_metric_name = 'loss'
objective_type = 'Minimize'
metric_definitions = [{'Name': 'loss',
                       'Regex': 'loss = ([0-9\\.]+)'}]
# Define hyperparameter ranges
hyperparameter_ranges = {
                            'learning_rate': ContinuousParameter(0.0001, 0.01), 
                            'dropout_rate': ContinuousParameter(0.3, 1.0),
                            'nw_depth': IntegerParameter(1, 4),

EniasCailliau

import sagemaker

bucket = sagemaker.Session().default_bucket() 
prefix = 'radix/mnist_fashion_tutorial' 

role = sagemaker.get_execution_role() 

import boto3
from time import gmtime, strftime
from sagemaker.tensorflow import TensorFlow

EniasCailliau

import os
import tensorflow as tf
from tensorflow.python.estimator.model_fn import ModeKeys as Modes
from sagemaker_tensorflow import PipeModeDataset
from tensorflow.contrib.data import map_and_batch

INPUT_TENSOR_NAME = 'inputs'
SIGNATURE_NAME = 'predictions'
PREFETCH_SIZE = 10
BATCH_SIZE = 128

EniasCailliau

convert_mnist_fashion_dataset(train_images, train_labels, 'train', 'data')
convert_mnist_fashion_dataset(test_images, test_labels, 'validation', 'data')

import sagemaker
bucket = sagemaker.Session().default_bucket() # Automatically create a bucket
prefix = 'radix/mnist_fashion_tutorial' # Subfolder prefix

s3_url = sagemaker.Session().upload_data(path='data', 
                                         bucket=bucket, 
                                         key_prefix=prefix+'/data/mnist')