Cloud ML Engine - Qwikstart

Slides for the talk given at Google Cloud Community New Delhi covering the basics of Cloud ML Engine.

Transcript

1. Alkesh Srivastava, Smalt and Beryl @_alkesh_ Cloud ML Engine: Qwikstart
2. What is Cloud ML?
3. Cloud ML 1. Hosted platform to run machine learning training jobs and predictions at scale. 2. Train and deploy machine learning models in the cloud 3. Part of the AI platform now
4. Why use Cloud ML?
5. Benefits of Cloud ML 1. Leverage powerful CPU/GPU for training 2. Utilize distributed training 3. Automatic resource provisioning 4. Automatic tuning of deep learning hyperparameters
6. Where does Cloud ML fit in?
8. Qwiklabs
9. About Qwiklabs 1. Hands-on learning platform for those who want to learn more about Google products 2. Provides a fun way to train and onboard people across all our products on Google Cloud
10. Cloud ML Engine: Qwik Start
11. Objectives of the lab 1. Create a TensorFlow training application and validate it locally. 2. Run your training job on a single worker instance in the cloud. 3. Deploy a model to support prediction. 4. Request an online prediction and see the response.
12. Before we begin
13. Basic terminology 1. Project - a logical container for your deployed models and jobs 2. Model - a logical container for individual versions of a ML solution 3. Jobs - Long-running operations that are processed asynchronously
14. 7 steps of machine learning 1. Gathering data 2. Preparing data 3. Choosing a model 4. Training 5. Evaluation 6. Hyperparameter tuning 7. Prediction
15. Wide and deep models
17. Demo
18. You can also try
19. Distributed training 1. set --scale-tier to any tier above BASIC 2. Specify OUTPUT_PATH to include the job name 3. Use gcloud ml-engine jobs submit training $JOB_NAME with --scale-tier to submit a training job in the cloud that uses multiple workers
20. Hyperparameter tuning 1. Hyperparameter configuration settings are stored in a YAML file named hptuning_config.yaml 2. Create a new job name and store the config file reference in a variable like HPTUNING_CONFIG=../hptuning_config. yaml
21. Alkesh Srivastava, Smalt and Beryl @_alkesh_ alkeshsrivastava.com Thank you!