# BOLT: An Automated Deep Learning Framework for Training and Deploying Large-Scale Search and Recommendation Models on Commodity CPU Hardware

Source: [BOLT: An Automated Deep Learning Framework for Training and Deploying Large-Scale Search and Recommendation Models on Commodity CPU Hardware](<https://www.sota2.com/research/paper/68fa70951e209131cea4bffb>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T03:48:30.799Z

Authors: Nicholas Meisburger, Vihan Lakshman, Benito Geordie, Joshua Engels, David Torres Ramos, Pratik Pranav, Benjamin Coleman, Benjamin Meisburger, Shubh Gupta, Yashwanth Adunukota, Tharun Medini, Anshumali Shrivastava

Published: 2023-03-30T22:03:43.000Z

## Abstract

Efficient large-scale neural network training and inference on commodity CPU hardware is of immense practical significance in democratizing deep learning (DL) capabilities. Presently, the process of training massive models consisting of hundreds of millions to billions of parameters requires the extensive use of specialized hardware accelerators, such as GPUs, which are only accessible to a limited number of institutions with considerable financial resources. Moreover, there is often an alarming carbon footprint associated with training and deploying these models. In this paper, we take a step towards addressing these challenges by introducing BOLT, a sparse deep learning library for training large-scale search and recommendation models on standard CPU hardware. BOLT provides a flexible, high-level API for constructing models that will be familiar to users of existing popular DL frameworks. By automatically tuning specialized hyperparameters, BOLT also abstracts away the algorithmic details of sparse network training. We evaluate BOLT on a number of information retrieval tasks including product recommendations, text classification, graph neural networks, and personalization. We find that our proposed system achieves competitive performance with state-of-the-art techniques at a fraction of the cost and energy consumption and an order-of-magnitude faster inference time. BOLT has also been successfully deployed by multiple businesses to address critical problems, and we highlight one customer case study in the field of e-commerce.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Node Classification on YelpChi Non-Homophilous Graph Benchmarks (test)](<https://www.sota2.com/research/sota/node-classification-on-yelpchi-non-homophilous-graph-benchmarks-test.md>): Node Classification; YelpChi Non-Homophilous Graph Benchmarks (test); ROC-AUC: 0.9318.

- [Personalized Recommendation on Movielens1M](<https://www.sota2.com/research/sota/personalized-recommendation-on-movielens1m.md>): Personalized Recommendation; Movielens1M; Inference Latency (ms): 1.

- [Personalized Recommendation on Amazon Games](<https://www.sota2.com/research/sota/personalized-recommendation-on-amazon-games.md>): Personalized Recommendation; Amazon Games; Recall@10: 13.4.

- [Personalized Recommendation on Netflix100M](<https://www.sota2.com/research/sota/personalized-recommendation-on-netflix100m.md>): Personalized Recommendation; Netflix100M; Recall@10: 6.61.

- [Personalized Recommendation on Amazon Games (Latency)](<https://www.sota2.com/research/sota/personalized-recommendation-on-amazon-games-latency.md>): Personalized Recommendation; Amazon Games; Latency (ms): 10.

- [Personalized Recommendation Latency on Netflix 100M](<https://www.sota2.com/research/sota/personalized-recommendation-latency-on-netflix-100m.md>): Personalized Recommendation; Netflix 100M; Latency (ms): 7.

- [Extreme Classification on Amazon-670k](<https://www.sota2.com/research/sota/extreme-classification-on-amazon-670k.md>): Extreme Classification; Amazon-670K; Accuracy: 34.8.

- [Node Classification on Penn94 Non-Homophilous (test)](<https://www.sota2.com/research/sota/node-classification-on-penn94-non-homophilous-test.md>): Node Classification; Penn94 Non-Homophilous Graph Benchmarks (test); Accuracy: 81.26.

- [Node Classification on Pokec non-homophilous large-scale (test)](<https://www.sota2.com/research/sota/node-classification-on-pokec-non-homophilous-large-scale-test.md>): Node Classification; Pokec non-homophilous large-scale (test); Accuracy: 0.7806.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2303.17727>)

[PDF](<https://arxiv.org/pdf/2303.17727.pdf>)

[Implementation](<https://github.com/ThirdAIResearch/BOLT_Benchmarks>)
