Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

NSML: Meet the MLaaS platform with a real-world case study

Hanjoo Kim, Minkyu Kim, Dongjoo Seo, Jinwoong Kim, Heungseok Park, Soeun Park, Hyunwoo Jo, KyungHyun Kim, Youngil Yang, Youngkwan Kim, Nako Sung, Jung-Woo Ha

TL;DR

NSML tackles dependency management and collaboration bottlenecks by delivering a full-stack MLaaS platform tailored for private clusters. It integrates resource scheduling, containerized environments, dataset and session collaboration, visualization, and serving APIs, with parallel hyperparameter tuning and leaderboard capabilities. The authors validate NSML through reproducible experiments on MNIST, CIFAR-100, and ImageNet, and through three hosted ML competitions with real-world use cases. The results demonstrate NSML's potential to support enterprise-scale ML workflows and enable commercialization of trained models.

Abstract

The boom of deep learning induced many industries and academies to introduce machine learning based approaches into their concern, competitively. However, existing machine learning frameworks are limited to sufficiently fulfill the collaboration and management for both data and models. We proposed NSML, a machine learning as a service (MLaaS) platform, to meet these demands. NSML helps machine learning work be easily launched on a NSML cluster and provides a collaborative environment which can afford development at enterprise scale. Finally, NSML users can deploy their own commercial services with NSML cluster. In addition, NSML furnishes convenient visualization tools which assist the users in analyzing their work. To verify the usefulness and accessibility of NSML, we performed some experiments with common examples. Furthermore, we examined the collaborative advantages of NSML through three competitions with real-world use cases.

NSML: Meet the MLaaS platform with a real-world case study

TL;DR

NSML tackles dependency management and collaboration bottlenecks by delivering a full-stack MLaaS platform tailored for private clusters. It integrates resource scheduling, containerized environments, dataset and session collaboration, visualization, and serving APIs, with parallel hyperparameter tuning and leaderboard capabilities. The authors validate NSML through reproducible experiments on MNIST, CIFAR-100, and ImageNet, and through three hosted ML competitions with real-world use cases. The results demonstrate NSML's potential to support enterprise-scale ML workflows and enable commercialization of trained models.

Abstract

The boom of deep learning induced many industries and academies to introduce machine learning based approaches into their concern, competitively. However, existing machine learning frameworks are limited to sufficiently fulfill the collaboration and management for both data and models. We proposed NSML, a machine learning as a service (MLaaS) platform, to meet these demands. NSML helps machine learning work be easily launched on a NSML cluster and provides a collaborative environment which can afford development at enterprise scale. Finally, NSML users can deploy their own commercial services with NSML cluster. In addition, NSML furnishes convenient visualization tools which assist the users in analyzing their work. To verify the usefulness and accessibility of NSML, we performed some experiments with common examples. Furthermore, we examined the collaborative advantages of NSML through three competitions with real-world use cases.

Paper Structure

This paper contains 27 sections, 8 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Platforms for machine learning tasks
  4. Cloud Platform
  5. Key features
  6. System Overview
  7. Resources and Management
  8. Locality-Aware scheduling and residual resource defragmentation
  9. Scheduler fault tolerance
  10. Session and Resource Monitoring
  11. Datasets and Collaboration
  12. User Interface
  13. Command line interface
  14. Web Interface and Visualization
  15. Serving API
  16. ...and 12 more sections

Figures (8)

  • Figure 1: High level overview of NSML
  • Figure 2: Dataset view in NSML web application
  • Figure 3: Example of client tool command executing sequence
  • Figure 4: Example of NSML analysis view: three sessions with different hyperparameters shown on NSML scalar plot. Top: The comparison part of each session. Bottom: Each plot of event data which users reported. The orange colored session is highlighted by the mouse on the 'test/accuracy' plot.
  • Figure 5: Example of NSML leaderboard
  • ...and 3 more figures