Dockerize Spago Self Contained ML & NLP Library & Deploy on Okteto Cloud Using Okteto Stack

Jan 31, 20210 likes141 views

This document discusses using Docker for machine learning and natural language processing projects. It introduces Spago, an open source machine learning library written in Go, and demos deploying a Spago project to Kubernetes using Okteto. The key points are: - Docker is useful for ML development as it allows packaging models and code into containers for scalable and reproducible deployments. - Spago is a pure Go library for NLP tasks like named entity recognition and question answering that aims to optimize CPU usage. - Okteto allows easily deploying Docker Compose stacks to Kubernetes with minimal configuration. - A live demo then shows deploying a Spago project to Kubernetes with Okteto.

Spago Self ConTained
ML & NLP Library & Deploy on Okteto

Agenda
For all level Audience ( Docker Knowledge + point )
Docker for Golang Developer
ML + NLP
Why Docker + ML
Live Demo !
2

“
3
Sangam Biradar
Docker Community Leader
Cloud Native Developer .
Github: Sangam14
Join Community http://discord.kubedaily.live/

Dockerize Spago Self Contained ML & NLP Library & Deploy on Okteto Cloud Using Okteto Stack

Why Docker + ML?
Docker is Not just in ML but across the entire development phase

What you should & shouldn’t include
in ML Development Container
⩥ Only the ML frameworks & Dependencies
⩥ ML frameworks and Dependencies & training
code ( scaling workload on cluster )
6

7
Spago
A Machine Learning Library Written in Pure Go
Designed to support Neural Architectures in NLP

- educational purpose
- production use
- High Quality Implementation
Purpose of spago
8

Where would I use SpaGo ?
9
⩥ Production NLP system
⩥ Required Optimization of
CPU interference
⩥ Relies on GONUM assembly
code with SIMD instruction
⩥ CGO is bottleneck , so there
are experiment with intel
MKL, Apache Arrow
⩥ Also experiment with sparse
model , LSH , etc
⩥ Training of No-very-deep model
⩥ The training seems less
interesting as its expensive for
modern model even with
GPUS/TPUs

Ready to use
⩥ Compatible with pytorch model
⩥ Transformer 50+ pretrained model in 100+ langauges
10

Use for “production level ml/nlp”
⩥ Train shallow models
⩥ fine tune deep model
⩥ inference
11
⩥ Train deep model
⩥ fine tune deep model
⩥ inference
Weight importer

Real time examples
⩥ Named Entities Recognition
⩥ Hugging Face Importer
⩥ Question Answering
⩥ Masked Language Model
12

Okteto – Kubernetes Developer platform
13

Docker Compose on kubernetes
⩥ okteto stack is easy way to deploy your docker
compose within secs .
15

16
Image name (eg. any image from
docker hub
name of your stack
build context
Override default
entry point
Port exposed by
each service
Resource quota
volume

This document discusses Okteto, a Kubernetes development platform that allows developers to build and test Kubernetes applications locally or in the cloud. It introduces Okteto Cloud, which provides free access to secure Kubernetes namespaces for remote development. The document demonstrates how to install Okteto CLI, configure access to an Okteto Cloud namespace, deploy sample applications, and get started with cloud native development on Okteto. It also discusses Helm and how it can be used to deploy and manage Kubernetes applications.

Docker + Tenserflow + GOlang - Golang singapore Meetupsangam biradar

This document discusses using TensorFlow with Golang for machine learning tasks like image recognition. It provides instructions for cloning a GitHub repository containing a sample project that uses a pre-trained TensorFlow model within a Golang application to classify images. The application is built as a Docker image to perform image recognition by taking URLs as arguments and returning potential labels and probabilities. The document also briefly mentions the possibility of training custom models from Golang in TensorFlow.

welcome to gopherlabs - why go (golang)?sangam biradar

Google ko: fast Kubernetes microservice development in Go - Sangam Biradar, E...sangam biradar

Cloud Native Okteto Cloud sangam biradar

The document provides instructions for getting started with Okteto Cloud, a platform for developing and deploying containerized applications. It summarizes how to install the Okteto CLI, configure access to an Okteto Cloud namespace, and deploy sample applications written in Go, Python, Node.js, and Ruby by applying Kubernetes manifest files and using Okteto commands. It also lists credentials and links for the service.

Multi-stage Docker builds to make building easy!Milindu Sanoj Kumarage

This document describes using Docker multi-stage builds to build and deploy a typical web application with a React frontend and Go backend. It outlines building each part (frontend and backend) in separate Docker images with the appropriate tools, then copying the artifacts to a final Nginx image for production deployment. This allows building each part in isolation without installing unnecessary tools in the final image, resulting in smaller Docker images for the application.

Docker based-Pipelines with CodefreshCodefresh

>>WATCH THE WEBINAR HERE: https://codefresh.io/docker-based-pipelines-with-codefresh/ Most people think that Docker adoption means deploying Docker images. In this webinar, we will see the alternative way of adopting Docker in a Continuous Integration Pipeline, by packaging all build tools inside Docker containers. This makes it very easy to use different tool versions on the same build and puts an end to version conflicts in build machines. We will use Codefresh as a CI/CD solution as it fully supports pipelines where each build step is running on its own container image. Sign up for FREE Codefresh account (120 builds/month) at Codefresh.io/codefresh-signup

Docker for local developmentAdam Štipák

Docker allows developers to create virtual environments on their local machines for testing and building projects. It provides consistent environments that are independent of the host machine, making it easy for teams to ensure their projects work the same on any machine. While Docker works natively on Linux, additional tools are required to use it on macOS, with some issues around shared folders that can be resolved through NFS mounting.

from docker run to docker compose (Webelement #36)Adam Štipák

Docker run allows for easy building and running of containers but has limitations around scalability and orchestration. Docker compose addresses these issues by defining and running multi-container Docker applications with YAML files, allowing services to be split into separate containers and together in a single command. Docker compose provides a way to define and share environments for development, testing and production.

funcs, func expressions, closure, returning funcs, recursion, the stack -goph...sangam biradar

The document provides an overview of functions in Go including function definitions, parameters and arguments, returning values, func expressions, closures, callbacks, recursion, and defer statements. It begins with basic concepts like defining functions and calling them, then covers more advanced topics like func expressions where a function is assigned to a variable, closures which allow functions to access variables from the enclosing scope, callbacks where functions are passed as arguments to other functions, and recursion where functions call themselves. Examples are provided for each concept using Go playground links. The document aims to explain how functions work and behave as first-class citizens in Go, providing a hands-on tutorial of key function-related ideas.

Unlimited Staging EnvironmentsCodefresh

Watch the webinar here: https://codefresh.io/unlimited-staging-environments-lp/ Sign up for a FREE Codefresh account today: https://codefresh.io/codefresh-signup/ How to run complete, disposable apps on Kubernetes or Staging and Development. See the full video webinar on our blog at https://codefresh.io/webinars/unlimited_staging_environments_webinar/

Gitlab ci-cdDan MAGIER

This document discusses GitLab Continuous Integration (GitLab CI/CD). It defines continuous integration, continuous delivery, and continuous deployment. It explains that GitLab CI/CD uses pipelines made up of stages and jobs to test, build, and deploy code. Pipelines are configured using a YAML file. Jobs run on GitLab runners, which can execute jobs locally or using Docker. Benefits of GitLab CI/CD include integrated pipelines, Docker/Kubernetes integration, and not requiring plugins. The downside is that it is only available within GitLab.

Multi-cloud CI/CD with failover powered by K8s, Istio, Helm, and Codefresh Codefresh

Codefresh is a Kubernetes-native CI/CD product used by 20,000 users globally. The document discusses using Codefresh to build a CI/CD pipeline across multiple Kubernetes clusters in different clouds or regions. This provides high availability and allows mitigating outages by failing over workloads between clusters. It also demonstrates using Istio for a canary deployment process, deploying a new version to a percentage of traffic initially and then routing all traffic to it if everything works as expected.

DevAssistant, Docker and YouBalaBit

This document discusses Docker, DevAssistant, and how they can help with development workflows. It summarizes that Docker allows for containerization of applications and their dependencies to enable portable deployment. While Docker simplifies development in some ways, properly setting up related containers and services can be complex. DevAssistant aims to address this by providing plugins to automate common development tasks like initializing projects, installing dependencies, and configuring container-based environments. The document demonstrates how DevAssistant can help simplify using Docker for development. It concludes with information on a competition to earn a free beer by following instructions using DevAssistant.

Continuous Delivery at Gogo with Spinnaker and ForemastN. Douglas Campbell

Automated Serverless Pipelines with #GitOps on CodefreshCodefresh

**Watch the full presentation here: https://codefresh.io/automated-serverless-pipelines-with-gitops-on-codefresh/ Dan Van Brunt introduces you to Serverless, talks about common misconceptions and challenges, and then demos how he uses the Serverless Framework effectively alongside containers. He shares some of the advanced pipelines he's developed so you can replicate his workflow without building a pipeline from scratch! Try Codefresh for FREE (120 builds/month) and get a free custom demo at Codefresh.io

Docker Best Practices WorkshopAhmed AbouZaid

A hands-on workshop that covers 18 best practices in 4 categories or in other words ✅️ Dos &  Don'ts. After a general introduction, we will have a look at the essential practices (aka must do), then move to the image practices, then we will go through the security practices, and finally, some general practices. Please note, this workshop assumes that you have a basic knowledge of Docker. Hands-on repo: https://github.com/aabouzaid/docker-best-practices-workshop

GitLab - Java User GroupPhilippWestphalen

This document summarizes a company's migration from Jenkins to GitLab CI/CD. They started by gathering information about GitLab's features and testing it on small projects. They migrated their main project in phases, first moving the repository to GitLab and then migrating the CI/CD pipeline. They took time to improve the pipeline by splitting jobs, changing the build tool to Gradle, using standard Docker images, and making it more efficient. Key learnings included planning the migration step-by-step while keeping infrastructure running, rethinking their DevOps strategy, and starting with a small project. GitLab suited them well with its powerful and customizable CI/CD solution providing a single home for their code, builds, and

Essential parts to implement own Ozone backendIgalia

This document provides instructions for implementing an Ozone backend using GTK on Chromium. It outlines the key entry points and classes that need to be implemented, including OzonePlatformGtk, GtkOzoneWindow, GtkScreen, and classes for the GPU side like SurfaceFactoryOzone and SurfaceOzoneCanvas. Implementing these classes and interfaces would allow Chromium to draw its content to a GTK window using the new Ozone/GTK backend.

The Self-Service Developer - GOTOCon CPHLaszlo Fogas

The Self-Service Developer In this talk, Laszlo revisits the primary motivations of why companies adopt container technologies in such a rapid pace and what became possible with the latest innovation in the container and CI/CD ecosystem. The talk iterates through the building blocks of self-service platforms. Platforms that allow developers to release software without asking for permission, or support from other teams or departments - so managers have fewer constraints to navigate, and instead of helping developers, ops people can focus on building platforms and resiliency into the system. The "self-service developer" vision is not only achievable by the big tech companies anymore, but increasingly becoming table stakes for every organization that produces software. During the talk, Laszlo shows a self-service platform in action featuring the latest CI/CD tools and the Kubernetes container orchestrator.

Ansible at glogsterRoman Komkov

This document summarizes how Ansible is used at Glogster. It discusses that Ansible was chosen over other tools like Puppet for its simplicity. Ansible is used to deploy RPM packages, manage services, and disable/enable monitoring across environments. It is also used with Jenkins for deployments and isolated tests. Ansible is ideal for security patches, package updates, and new server provisioning via roles, tags, and replacing tools like cssh/pssh. The document encourages keeping everything in Git, avoiding redundant logic, using tags, and treating Ansible as both tests and documentation for servers. It also discusses desires like using Docker, wrapping infrastructure in a single playbook, using dynamic inventory, and keeping facts in cache.

How to Become DevOpsOpsta

Agile NCR 2013- Shekhar Gulati - Open shift platform-for-rapid-and-agile-deve...AgileNCR2013

OpenShift is a platform as a service (PaaS) by Red Hat that allows developers to rapidly develop and deploy applications in the cloud. The presentation demonstrates how to use OpenShift through its web console and command line tools to create Java and MySQL applications integrated with Jenkins for continuous integration. It also shows how to install the code quality tool Sonar and agile issue tracking tool YouTrack on OpenShift. The key benefits of OpenShift are that it allows developers to focus on coding while handling deployment, scaling, and infrastructure management.

Accelerate your business and reduce cost with OpenStackOpsta

Microservices in GolangMo'ath Qasim

Every single developer heard this say more than one time during his professional career, 'KIS' which means 'Keep It Simple', however with the excessive evolvement of the web computing and the continues rise of the distributed systems, the complexity of the systems architecture increases. Building a server enterprise application which is highly scalable with high and reliable availability in addition to great performance which consists of set of independent deployable oriented services can be achieved by Microservice architecture. This session introduces briefly what is the Microservice Architecture and how to build it using one of the most popular server programming languages which is called Go or Golang.

Istio + Helm + Canary WebinarCodefresh

**WATCH THE WEBINAR HERE: https://codefresh.io/canary_helm_lp/ Sign up for a FREE Codefresh account today: https://codefresh.io/codefresh-signup/ Istio is unlocking some amazing ways to handle deployment networking for Kubernetes, but what about when you add in the wrinkle of Helm? In this webinar, Kubernaut Dan Garfield will show how to bring all these technologies together. Who knew Helm would pair so well with Istio?

Gitlab ci, cncf.skJuraj Hantak

This document provides an introduction to Gitlab CI and continuous integration/continuous delivery (CI/CD) workflows. It discusses DevOps practices and the benefits of Gitlab CI. It then covers how to set up Gitlab runners, write a basic Gitlab CI configuration file, define jobs, stages, variables and environments. The document demonstrates concepts like Docker integration, artifacts, auto and manual deployments, and stopping deployments. It concludes with a live demo of a Gitlab CI configuration.

Docker 101Hung-Che Lo

Terrascan - Cloud Native Security Tool sangam biradar

XCloudLabs- AWS Overview sangam biradar

This document provides an overview of AWS networking services including Virtual Private Cloud, Amazon Route 53, AWS Direct Connect, VPN, and Elastic Load Balancing. It describes each service's purpose such as Virtual Private Cloud allowing users to launch AWS resources in a virtual private network and Amazon Route 53 providing scalable and available cloud DNS. The document also defines networking terminology like scalability, fault tolerance, elasticity, durability, and availability.

More Related Content

What's hot (20)

from docker run to docker compose (Webelement #36)Adam Štipák

funcs, func expressions, closure, returning funcs, recursion, the stack -goph...sangam biradar

Unlimited Staging EnvironmentsCodefresh

Gitlab ci-cdDan MAGIER

Multi-cloud CI/CD with failover powered by K8s, Istio, Helm, and Codefresh Codefresh

DevAssistant, Docker and YouBalaBit

Continuous Delivery at Gogo with Spinnaker and ForemastN. Douglas Campbell

Automated Serverless Pipelines with #GitOps on CodefreshCodefresh

Docker Best Practices WorkshopAhmed AbouZaid

GitLab - Java User GroupPhilippWestphalen

Essential parts to implement own Ozone backendIgalia

The Self-Service Developer - GOTOCon CPHLaszlo Fogas

Ansible at glogsterRoman Komkov

How to Become DevOpsOpsta

Agile NCR 2013- Shekhar Gulati - Open shift platform-for-rapid-and-agile-deve...AgileNCR2013

Accelerate your business and reduce cost with OpenStackOpsta

Microservices in GolangMo'ath Qasim

Istio + Helm + Canary WebinarCodefresh

Gitlab ci, cncf.skJuraj Hantak

Docker 101Hung-Che Lo

from docker run to docker compose (Webelement #36)Adam Štipák

funcs, func expressions, closure, returning funcs, recursion, the stack -goph...sangam biradar

Unlimited Staging EnvironmentsCodefresh

Gitlab ci-cdDan MAGIER

Multi-cloud CI/CD with failover powered by K8s, Istio, Helm, and Codefresh Codefresh

DevAssistant, Docker and YouBalaBit

Continuous Delivery at Gogo with Spinnaker and ForemastN. Douglas Campbell

Automated Serverless Pipelines with #GitOps on CodefreshCodefresh

Docker Best Practices WorkshopAhmed AbouZaid

GitLab - Java User GroupPhilippWestphalen

Essential parts to implement own Ozone backendIgalia

The Self-Service Developer - GOTOCon CPHLaszlo Fogas

Ansible at glogsterRoman Komkov

How to Become DevOpsOpsta

Agile NCR 2013- Shekhar Gulati - Open shift platform-for-rapid-and-agile-deve...AgileNCR2013

Accelerate your business and reduce cost with OpenStackOpsta

Microservices in GolangMo'ath Qasim

Istio + Helm + Canary WebinarCodefresh

Gitlab ci, cncf.skJuraj Hantak

Docker 101Hung-Che Lo

More from sangam biradar (14)

Terrascan - Cloud Native Security Tool sangam biradar

XCloudLabs- AWS Overview sangam biradar

Rustlabs Quick Startsangam biradar

This document provides an introduction to the Rust programming language. It discusses Rust's memory safety features, variable bindings, functions, control flow statements like if/else and loops, data types like tuples and vectors, and borrowing rules. It also covers Rust concepts like ownership, slices, pattern matching, and destructuring. Examples are provided to demonstrate various Rust language features.

Okteto For Kubernetes Developer :- Container Camp 2020 sangam biradar

Okteto is a development platform that allows developers to build and test Kubernetes applications directly in Kubernetes clusters without disrupting their development workflow. It works with local Kubernetes clusters like Minikube as well as remote clusters on cloud providers. Developers can use Okteto to get production-like environments in sandboxed Kubernetes namespaces to develop and test their code directly in a Kubernetes cluster.

5 cool ways to get started with Cloud Native Development ( with Okteto)sangam biradar

This document introduces 5 cool ways to get started with cloud native development: 1) Okteto Stacks for developing with docker-compose in the cloud native world, 2) Okteto Push to push code to Kubernetes in seconds, 3) VS Code Remote for remote development environments in Kubernetes from VS Code, 4) Okteto Actions to build, preview, and ship cloud native apps from Github, and 5) OpenFaaS + Okteto for the easiest way to develop and debug serverless functions. It provides demo links and invites stars, feedback, issues and pull requests on Github.

kikstart journey of Golang with Hello world - Gopherlabs sangam biradar

Decision making - for loop , nested loop ,if-else statements , switch in goph...sangam biradar

This document discusses decision making in Golang. It provides an overview of loops including for, while, break, continue, and nested loops. It also covers conditionals such as if, else if, else, switch statements, and logical operators. Code examples are provided for each concept via links to an online Golang playground. The author is identified as Sangam Biradar, a Docker community leader who writes tutorials on Golang.

Types - slice, map, new, make, struct - Gopherlabs sangam biradar

This document provides an overview of Go programming concepts including slices, maps, structs, make, and new. It includes links to interactive code examples demonstrating how to use slices to store and access elements, maps to associate keys with values, and structs to group related data. The key differences between make and new are explained, where make is used to initialize slices, maps, and channels, and new returns a pointer to a newly allocated zero value.

Rabncher Meetup India , Lightweight Kubernetes Development with K3s, k3os and...sangam biradar

TensorFlow, Docker & GoLang - All for Image Rekognition Sangam Biradar(Engine...sangam biradar

Introducing Pico - A Deep Learning Platform using Docker & IoT - Sangam Biradarsangam biradar

This document discusses simplifying real-time data analytics with Docker, IoT, and AWS Rekognition Service. It introduces Sangam Biradar and their experience with Docker, Golang, Kubernetes, cloud computing, and more. It then discusses popular IoT boards like Raspberry Pi and integrating a Pi with a Celestron telescope. The rest of the document focuses on using Docker and related technologies to build a low-cost CCTV security system using object detection and analytics on video streams from a Raspberry Pi camera.

September 7, 2019 Cloud Native and Containerisation (Joint Meetup with Docke...sangam biradar

The document summarizes a presentation given by Sangam Biradar on Docker internals. It introduces the key building blocks of containers like namespaces, control groups, copy-on-write storage and union filesystems. It then explains the container runtime and demonstrates how to create a minimal Docker container using Golang. The presentation outlines the key components and how they work together to provide operating-system-level virtualization and isolation for containers.

Implementing Lightweight Kubernetes(K3s) on Raspberry Pi Stack - Sangam Biradarsangam biradar

Docker on IOT - Dockercon19 SFO Recap & Announcements, Bangaloresangam biradar

This document discusses using Docker on IoT devices like Raspberry Pi. It provides steps to set up a Docker Swarm cluster on Raspberry Pi and create a Docker container to blink an LED connected to a Raspberry Pi. Some key points covered include why IoT needs Docker due to limited hardware resources, setting up the Docker Swarm cluster, creating a Dockerfile and Python script to blink an LED, building the Docker image, and running the container on Raspberry Pi. Examples codes for the Dockerfile, Python script and commands to build/run the container are also included.