# [Building and Benchmarking DLRM on Arm Neoverse V2 with MLPerf](https://learn.arm.com/learning-paths/servers-and-cloud-computing/dlrm/)

## In this learning path

- [Introduction](https://learn.arm.com/learning-paths/servers-and-cloud-computing/dlrm/)
- [Overview and setup](https://learn.arm.com/learning-paths/servers-and-cloud-computing/dlrm/1-overview/)
- [Download model weights and data](https://learn.arm.com/learning-paths/servers-and-cloud-computing/dlrm/2-download-model/)
- [Run the benchmark](https://learn.arm.com/learning-paths/servers-and-cloud-computing/dlrm/3-run-benchmark/)
- [Next Steps](https://learn.arm.com/learning-paths/servers-and-cloud-computing/dlrm/_next-steps/)

## About this Learning Path

| Skill level:       | Introductory        |
|---------------------|---------------------|
| Reading time:       | 1 hr 30 min         |
| Last updated:       | 31 Jul 2026         |

### Authors:
- Phalani Paladugu, Arm [GitHub](https://github.com/phalani-paladugu) [LinkedIn](https://linkedin.com/in/phalani-paladugu)
- Annie Tallund, Arm [GitHub](https://github.com/annietllnd) [LinkedIn](https://linkedin.com/in/annietallund)
- Pareena Verma, Arm [GitHub](https://github.com/pareenaverma) [LinkedIn](https://linkedin.com/in/pareena-verma-7853607)

### Arm IP:
[Neoverse](https://support.arm.com/?tab=compute-ip&Product%20Type=Infrastructure%20Processors)

### Tags:
- [Performance and Architecture](https://learn.arm.com/tag/performance-and-architecture)
- [AWS](https://learn.arm.com/tag/aws)
- [Google Cloud](https://learn.arm.com/tag/google-cloud)
- [Linux](https://learn.arm.com/tag/linux)
- [Docker](https://learn.arm.com/tag/docker)
- [MLPerf](https://learn.arm.com/tag/mlperf)

### Who is this for?
This is an introductory topic for software developers who want to set up a pipeline in the cloud for recommendation models. You'll build and run the Deep Learning Recommendation Model (DLRM) and benchmark its performance using MLPerf and PyTorch.

### What will you learn?
Upon completion of this Learning Path, you will be able to:
- Build the Deep Learning Recommendation Model (DLRM).
- Run a modified performant DLRMv2 benchmark and inspect the results.

### Prerequisites
Before starting, you will need the following:
- Any [Arm-based instance](https://learn.arm.com/learning-paths/servers-and-cloud-computing/csp/) from a cloud service provider (CSP), or an on-premise Arm server with at least 400GB of RAM and 800 GB of disk space.

### Summary
You’ll prepare an Arm Neoverse V2 server, download the DLRM dataset and model weights with `rclone`, and run a modified MLPerf DLRM benchmark. You’ll create dedicated directories, clone the provided repository, use PyTorch 2.9.0+cpu with Arm-focused optimizations, and execute the benchmark. You’ll then inspect the output to confirm a successful run and review the results.

### Frequently asked questions

#### How do I know I’ve installed `rclone` correctly, and do I need to configure it before downloading?
After the install script reports that `rclone` installed successfully, run `rclone config` before downloading the data and model weights.

#### Which directories should contain the downloaded dataset and model weights?
Create data and model directories under your home directory, for example `$HOME/data` and `$HOME/model`. Download the dataset into `data` and the model weights into `model`.

#### Which PyTorch build does the benchmark use?
The benchmark uses PyTorch 2.9.0+cpu with optimizations for recommendation models on Arm. Use the build referenced by the provided scripts.

#### How do I know the benchmark finished correctly, and where do I view the results?
The run prints benchmark progress and produces results you can inspect after completion. Review the script output and confirm that the run completed without errors.

#### What should I check if the download or benchmark fails partway through?
Verify the instance meets the stated resource requirements and that sufficient RAM and disk space are available. Also confirm that the dataset and model weights reached the expected directories before rerunning.