NVIDIA Merlin™ is an open-source framework for building high-performing recommender systems at scale.
Designed for Recommender Workflows
NVIDIA Merlin empowers data scientists, machine learning engineers, and researchers to build high-performing recommenders at scale. Merlin includes libraries, methods, and tools that streamline the building of recommenders by addressing common preprocessing, feature engineering, training, inference, and deploying to production challenges. Merlin components and capabilities are optimized to support the retrieval, filtering, scoring, and ordering of hundreds of terabytes of data, all accessible through easy-to-use APIs. With Merlin, better predictions, increased click-through rates, and faster deployment to production are within reach.
Interoperable End-to-End Solution
NVIDIA Merlin, as part of NVIDIA AI, advances our commitment to support innovative practitioners doing their best work. As an end-to-end solution, NVIDIA Merlin components are designed to be interoperable within existing recommender workflows that utilize data science, machine learning (ML), and deep learning (DL) on CPUs or GPUs. Data scientists, ML engineers, and researchers are able to use single or multiple components to accelerate the entire recommender pipeline—from ingesting, training, inference, to deploying to production. NVIDIA Merlin's open-source components simplify building and deploying a production-quality pipeline.
Merlin Models is a library that provides standard models for recommender systems and high quality implementations from ML to more advanced DL models on CPUs and GPUs. Train models for retrieval and ranking within 10 lines of code.
Merlin NVTabular is a feature engineering and preprocessing library designed to effectively manipulate terabytes of recommender system datasets and significantly reduce data preparation time.
Merlin Transformers4Rec is a library that streamlines the building of pipelines for session-based recommendations. The library makes it easier to explore and apply popular transformers architectures when building recommenders.
Merlin Distributed Training
Merlin provides support for distributed training across multiple GPUs. Components include Merlin SOK (SparseOperationsKit) and Merlin Distributed Embeddings (DE). TensorFlow (TF) users are empowered to use SOK (TF 1.x) and DE (TF 2.x) to leverage model parallelism to scale training.
Built on NVIDIA AI
NVIDIA AI empowers millions of hands-on practitioners and thousands of companies to use the NVIDIA AI Platform to accelerate their workloads. NVIDIA Merlin, is part of the NVIDIA AI Platform. NVIDIA Merlin was built upon and leverages additional NVIDIA AI software within the platform.
RAPIDS is a suite of open source software libraries and APIs that enables end-to-end data science and analytics pipelines entirely on GPUs.
cuDF i is a Python GPU DataFrame library for loading, joining, aggregating, filtering, and manipulating data.
NVIDIA Triton Inference Server
Take advantage of NVIDIA Triton™ Inference Server to run inference efficiently on GPUs by maximizing throughput with the right combination of latency and GPU utilization.
Recommender Systems, Not Just Recommender Models
Learn about the 4 stages of recommender systems pattern that covers the majority of modern recommenders today.
GTC Spring 2022 Keynote: NVIDIA Merlin
Watch as NVIDIA CEO Jensen Huang discusses how recommenders personalize the internet and highlights NVIDIA Merlin use cases from Snap and Tencent's WeChat.
Optimizing ML Platform with NVIDIA Merlin
Discover how Meituan optimizes their ML Platform using NVIDIA Merlin.
Take this survey to share some information about your recommender pipeline and influence the NVIDIA Merlin roadmap.