NSML: A Machine Learning Platform That Enables You to Focus on Your Models

2017/12/16 by Nako Sung, Minkyu Kim, Sung, Nako +19 · 8 citations
Computer Science · #Data Stream Mining Techniques #Distributed #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning and Data Classification #Parallel #and Cluster Computing (cs.DC)

paper · pdf · doi:10.48550/arxiv.1712.05902

openalex publication_date 2017/12/16 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Machine learning libraries such as TensorFlow and PyTorch simplify model implementation. However, researchers are still required to perform a non-trivial amount of manual tasks such as GPU allocation, training status tracking, and comparison of models with different hyperparameter settings. We propose a system to handle these tasks and help researchers focus on models. We present the requirements of the system based on a collection of discussions from an online study group comprising 25k members. These include automatic GPU allocation, learning status visualization, handling model parameter snapshots as well as hyperparameter modification during learning, and comparison of performance metrics between models via a leaderboard. We describe the system architecture that fulfills these requirements and present a proof-of-concept implementation, NAVER Smart Machine Learning (NSML). We test the system and confirm substantial efficiency improvements for model development.

Citations

• ImageNet Large Scale Visual Recognition Challenge
• MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
• Densely Connected Convolutional Networks
• YouTube-8M: A Large-Scale Video Classification Benchmark
• A Structured Self-attentive Sentence Embedding
• Deep Voice 2: Multi-Speaker Neural Text-to-Speech
• YOLO9000: Better, Faster, Stronger

Cited by

• StyleKeeper: Prevent Content Leakage using Negative Visual Query Guidance
• Toward Interpretable Music Tagging with Self-Attention
• Boosting Active Learning for Speech Recognition with Noisy Pseudo-labeled Samples
• Designing a Minimal Retrieve-and-Read System for Open-Domain Question Answering
• NeuralWOZ: Learning to Collect Task-Oriented Dialogue via Model-Based Simulation
• Subword Language Model for Query Auto-Completion
• Exploiting Spatial Dimensions of Latent in GAN for Real-time Image Editing
• DialogWAE: Multimodal Response Generation with Conditional Wasserstein Auto-Encoder

Related

• NSML: Meet the MLaaS platform with a real-world case study
• Snap ML: A Hierarchical Framework for Machine Learning
• Superpipeline: A Universal Approach for Reducing GPU Memory Usage in Large Models
• Talaria: Interactively Optimizing Machine Learning Models for Efficient Inference
• MLModelScope: A Distributed Platform for Model Evaluation and Benchmarking at Scale