Data scientists and machine learning engineers use many methods, techniques, and tools to prep, build, train, deploy, and optimize their machine learning models. While technical leads cite the importance of leveraging open source software for recommender team workflows, the majority of popular machine learning methods, libraries, and frameworks are not designed to support and accelerate recommender workflows.
NVIDIA Merlin is designed to streamline recommender workflows. The latest update includes Transformers4Rec, a new library that wraps HuggingFace Transformer Architectures to build pipelines for session-based recommendations. It also adds SparseOperationsKit (SOK), a new Python package that supports sparse training and inference with Deep Learning (DL).
This latest release reaffirms the commitment of NVIDIA to help machine learning engineers and data scientists develop and optimize their recommender systems—with open source canonical building blocks.
Merlin Transformers4Rec, designed for recommenders and solving cold-start problems
Recommender methods popularized in mainstream media often rely upon long-term user profiles or lifetime user behavior. Yet, ecommerce and media companies acquiring new ongoing active users must provide relevant recommendations to first-time and early-visit users. Relevant recommendations enable increased user engagement, retention, and conversion to subscription services.
Utilizing session-based recommenders with Transformers4Rec, data scientists and machine learning engineers are able to solve the cold-start problem by leveraging contextual and recent user interactions to predict a user’s next action and provide relevant recommendations. The NVIDIA Merlin team designed Transformers4Rec to be used as a standalone solution or within an ensemble of recommendation models.
SparseOperationsKit, sparse training, and inference with deep learning
Recommender teams that work with massive datasets benefit from using deep learning (DL) recommenders. Merlin HugeCTR is a DL training framework designed for recommender systems and the latest update includes SOK, a new open source Python package that supports sparse training and inference.
It is also compatible with DL frameworks including TensorFlow. SOK provides embedding model parallelism functionality to use GPUs, including scaling from a single GPU to multiple GPUs. Most common DL frameworks do not support model-parallelism, which makes it challenging to use all available GPUs in a cluster. Yet, SOK being compatible with DL frameworks, including TensorFlow, helps fill that void.
Download and try NVIDIA Merlin
The latest update to NVIDIA Merlin, including Transformers4Rec and SOK, strengthens streamlining and accelerating recommender workflows with open-source interoperability and performance enhancements.
For more information about the latest release download NVIDIA Merlin today.