title	theme	highlightTheme	separator	verticalSeparator	revealOptions
Google Colab & AI Notebook	solarized	solarized-dark	---	--	transition transitionSpeed controls fade default false

Google Colaboratory

& AI Notebook

Data Science in the cloud, the easy way

Toulouse Data Science #38 - June 18th 2019

Florient CHOUTEAU

--

about.md

• ML Engineer @ Airbus Defence and Space (Space Systems)

• Training Neural Networks on remote sensing imagery since 2016

  • Delair, Magellium & Airbus Intelligence (w/ Jeff, spoilers), Airbus DS...

  • torch, tf, keras, pytorch, ...

  • a lot of time spent installing instances

• Contact: @foxchouteau or on Slack

--

Who started learning data science recently ?

--

Who works in data science ?

--

Who teaches data science classes ?

--

TL;DR

• easy access to configured development environment for ML

• from Google but not limited to their tech

• jupyter-based products

• one free, one paid: different use cases, similar principles

--

Disclaimer

This talk is not sponsored by Google ;)

There may be better alternatives: Feel free to comment after :)

---

Colaboratory

https://colab.research.google.com

--

--

WTF is... Google Colab ?

• Jupyter Notebook + Google Drive

• Full python data science environment

• 12h max session lifetime

--

Is it for YOU ?

• Students, people learning ML/DS

• Teachers, share courses, get assignments

• Quick experiments / sharing

--

Nice features

• Can use your data: gdrive, gsheet, local filesystem

• Jupyter-based: All the power of interactive & visualisations

• You can apt-get and pip install what you need

--

Nicer features

• GPU ! (Nvidia Tesla T4, 16 GB GPU RAM = 3000$)

• Collaboration ! (share and co-edit notebooks)

• Open notebook from github to colab !

--

Demo Time

• End-to-end training w/ GPU. pytorch and pytorch-ignite

• Notebook on github, Data on Google Drive

• https://colab.research.google.com

--

Limitations

• Long calculations w/ guarantees (you can checkpoint your models on colab though)

• Code syncing / huge codebase & huge datasets

• Full control over installation and data

---

GCP Deep Learning VM / AI Platform Notebook

https://cloud.google.com/deep-learning-vm/

--

Google Cloud Platform

• Cloud Provider, very nice VM instances options

• 300$ free, paid for GPU and unlocked bandwidth

• Rather easy to use for ML / DS

--

WTF is... AI Platform Notebook ?

• Pre configured paid Cloud Virtual Machines (Google Compute Engine)

• With jupyter lab auto launched & ready

• Papermill pre installed for scheduling

--

Available configurations

--

2 different workflows

1. Jupyter only ("AI Notebook")

2. Pre-configured instance for Data Science ("Deep Learning VM")

--

Demo 1: "AI Platform Notebook"

• Creating an instance

• Connecting to jupyter lab (with or without ssh !)

https://console.cloud.google.com

--

Demo 2: "Deep Learning VM"

• Using the DL VM as a preconfigured headless code runner

• Executing a notebook on a deep-learning-vm

• Install this: https://github.com/gclouduniverse/gcp-notebook-executor

INPUT_NOTEBOOK="gs://{your-storage}/ai-notebook-demo.ipynb"
GCP_BUCKET="gs://{your-storage}/runs"
IMAGE_FAMILY_NAME="pytorch-latest-gpu"
INSTANCE_TYPE="n1-standard-8"
GPU_TYPE="k80"
GPU_COUNT=1
ZONE="europe-west1-b"

execute_notebook -i "${INPUT_NOTEBOOK}" \
                 -o "${GCP_BUCKET}" \
                 -f "${IMAGE_FAMILY_NAME}" \
                 -t "${INSTANCE_TYPE}" \
                 -z "${ZONE}" \
                 -g "${GPU_TYPE}" \
                 -c "${GPU_COUNT}"

--

Advanced Usage (not covered here)

• Extensive tutorial

• Use "preemptible" (spot in AWS terminology)*

• CLI creation for more customization

*5x less expensive, run only 24h

---

Conclusion

--

TL;DR (bis)

Google Colab	Google AI Notebook
Learn, experiment	Can scale compute
Single notebook / Clone from github	Upload own code
Simple jupyter env.	Full jupyter lab or SSH access
Data from anywhere / google drive	Fully owned cloud environment
Short runtimes	Cheap 1d runtimes or arbitrary runtimes
**free**	**[paid](https://cloud.google.com/compute/pricing)** (by minute of computing + storage)

--

Alternatives

• Kaggle Kernels: for kaggle, colab, free, 9h, P100

• Amazon Sagemaker: can someone tell me about it ?

• A lot of smaller entities... floydhub...

• Build your own machine ? opinion: last step for individual use (be sure of what you need !)

--

Thank you !