{"id":20326973,"url":"https://github.com/habibslim/distdnns","last_synced_at":"2026-06-07T17:31:47.991Z","repository":{"id":106054738,"uuid":"326426221","full_name":"HabibSlim/DistDNNs","owner":"HabibSlim","description":"Speeding up DNN training for image classification, with OpenMPI","archived":false,"fork":false,"pushed_at":"2021-10-27T09:18:22.000Z","size":77294,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-01-14T14:45:51.191Z","etag":null,"topics":["computer-vision","distributed-learning","machine-learning","openmpi"],"latest_commit_sha":null,"homepage":"","language":"C++","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/HabibSlim.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2021-01-03T14:29:11.000Z","updated_at":"2021-10-27T09:18:28.000Z","dependencies_parsed_at":null,"dependency_job_id":"6d9532d3-c492-47d5-b2a8-72b51550c712","html_url":"https://github.com/HabibSlim/DistDNNs","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HabibSlim%2FDistDNNs","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HabibSlim%2FDistDNNs/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HabibSlim%2FDistDNNs/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HabibSlim%2FDistDNNs/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/HabibSlim","download_url":"https://codeload.github.com/HabibSlim/DistDNNs/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":241836137,"owners_count":20028146,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["computer-vision","distributed-learning","machine-learning","openmpi"],"created_at":"2024-11-14T19:46:02.476Z","updated_at":"2026-06-07T17:31:47.981Z","avatar_url":"https://github.com/HabibSlim.png","language":"C++","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n\u003ch1 align=\"center\"\u003e\n    Speeding up DNN training in distributed environments\n\u003c/h1\u003e\n\n[![C++](https://img.shields.io/badge/C++-17-red?logo=c%2B%2B\u0026logoColor=white)](https://en.wikipedia.org/wiki/C%2B%2B17)\n[![Eigen](https://img.shields.io/badge/Eigen-3.0-brown?logo=Eigen\u0026logoColor=white)](http://eigen.tuxfamily.org/)\n[![OpenMPI](https://img.shields.io/badge/OpenMPI-2.1.1-blue?logo=Eigen\u0026logoColor=white)](https://www.open-mpi.org/)\n\n\\[[Report](./report.pdf)\\]\n\n\u003cp align=\"center\"\u003e\n     \u003cimg src=\"./main_fig.svg\" width=500px/\u003e\n\u003c/p\u003e \n\u003c/div\u003e\n\n\n## Summary\n\n* [Introduction](#introduction)\n* [Running](#running)\n* [References](#references)\n\n## Introduction\n\nThis repository contains a minimal framework to quickly prototype deep architectures and facilitate weight and gradients sharing among processing nodes.\n\nWe introduce and implement parallel DNN optimization algorithms and conduct a (hopefully complete) benchmark of the different methods, evaluated on the MNIST and Fashion-MNIST datasets.\nFor a full description of the project, you can check out the [project report](./report.pdf)!\n\n## Running\n\nThis project uses the following dependencies:\n- Eigen3, for basic matrix operations (also included in the repository)\n- OpenMPI 2.1.1\n\nThe list of experiments available is as follows:\n- `param_avg`: Weight averaging algorithm described in the report.\n- `parallel_sgd`: Gradient averaging algorithm described in the report.\n- `w_param_avg`: Weighted parameter averaging algorithm described in the report.\n\nTo compile and run the experiments, from the root directory:\n\n```bash\nmake experiment_name\n```\n\nAnd then, for an MPI experiment:\n\n```bash\nmpiexec -n n_cores runmpi -options\n```\n\nParameters are as follows. For all experiments, the following arguments are available:\n- `-batch_size`: Size of each batch\n- `-eval_acc`  : To be set to `1` if validation accuracies must be evaluated, `0` otherwise (in which case epoch durations are logged instead).\n- `-n_epochs`  : Total number of epochs\n\nSpecifically to the following methods, additional parameters are available\n\n**param_avg, w_param_avg**\n- `-avg_freq`: Weight averaging frequency (in epochs).\n\n**w_param_avg**\n- `-lambda`: Value of the lambda parameter (integer, divided by 100).\n\n\n## References\n1. [Ben-Nun et al., 2018] Demystifying parallel and distributed deep learning: An in-depth concurrency analysis\n2. [Ericson et al., 2017] On the performance of network parallel training in artificial neural networks\n3. [Han Xiao et al., 2017] Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhabibslim%2Fdistdnns","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fhabibslim%2Fdistdnns","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhabibslim%2Fdistdnns/lists"}