{"data":{"slug":"horovod-horovod","name":"horovod","tagline":"Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.","github_url":"https://github.com/horovod/horovod","owner":"horovod","repo":"horovod","owner_avatar_url":"https://avatars.githubusercontent.com/u/46361271?v=4","primary_language":"Python","stars":14695,"forks":2235,"topics":["baidu","deep-learning","deeplearning","keras","machine-learning","machinelearning","mpi","mxnet","pytorch","ray","spark","tensorflow","uber"],"archived":true,"github_pushed_at":"2026-07-29T17:27:00+00:00","maintenance_label":"Archived","url":"https://www.graphcanon.com/tools/horovod-horovod","markdown_url":"https://www.graphcanon.com/tools/horovod-horovod.md","api_url":"https://www.graphcanon.com/api/graphcanon/tools/horovod-horovod","graph_url":"https://www.graphcanon.com/api/graphcanon/graph?tool=horovod-horovod","description":"Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.","homepage_url":"http://horovod.ai","license":"Other","open_issues":406,"watchers":322,"ai_summary":"Horovod is a distributed deep learning training framework designed to make distributed deep learning fast and easy to use. It supports popular machine learning frameworks like TensorFlow, Keras, PyTorch, and Apache MXNet, facilitating multi-GPU and multi-node scaling.","readme_excerpt":".. raw:: html\n\n
\n\n.. inclusion-marker-start-do-not-remove\n\n|\n\nHorovod is a distributed deep learning training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.\nThe goal of Horovod is to make distributed deep learning fast and easy to use.\n\n\n.. raw:: html\n\n
\n\n\nHorovod is hosted by the `LF AI & Data Foundation `_ (LF AI & Data). If you are a company that is deeply\ncommitted to using open source technologies in artificial intelligence, machine, and deep learning, and want to support\nthe communities of open source projects in these domains, consider joining the LF AI & Data Foundation. For details\nabout who's involved and how Horovod plays a role, read the Linux Foundation `announcement `_.\n\n|\n\n.. contents::\n\n|\n\nDocumentation\n-------------\n\n- `Latest Release `_\n- `master `_\n\n|\n\nWhy Horovod?\n------------\nThe primary motivation for this project is to make it easy to take a single-GPU training script and successfully scale\nit to train across many GPUs in parallel. This has two aspects:\n\n1. How much modification does one have to make to a program to make it distributed, and how easy is it to run it?\n2. How much faster would it run in distributed mode?\n\nInternally at Uber we found the MPI model to be much more straightforward and require far less code changes than previous\nsolutions such as Distributed TensorFlow with parameter servers. Once a training script has been written for scale with\nHorovod, it can run on a single-GPU, multiple-GPUs, or even multiple hosts without any further code changes.\nSee the `Usage <#usage>`__ section for more details.\n\nIn addition to being easy to use, Horovod is fast. Below is a chart representing the benchmark that was done on 128\nservers with 4 Pascal GPUs each connected by RoCE-capable 25 Gbit/s network:\n\n.. image:: https://user-images.githubusercontent.com/16640218/38965607-bf5c46ca-4332-11e8-895a-b9c137e86013.png\n :alt: 512-GPU Benchmark\n\nHorovod achie","github_created_at":"2017-08-09T19:39:59+00:00","created_at":"2026-07-11T23:23:08.794809+00:00","updated_at":"2026-08-03T00:02:03.825543+00:00","categories":[{"slug":"model-training","name":"Model Training","url":"https://www.graphcanon.com/categories/model-training","markdown_url":"https://www.graphcanon.com/categories/model-training.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/model-training"}],"tags":[{"slug":"deep-learning","name":"deep-learning"},{"slug":"distributed-training","name":"distributed-training"},{"slug":"keras","name":"keras"},{"slug":"mxnet","name":"mxnet"},{"slug":"pytorch","name":"pytorch"},{"slug":"tensorflow","name":"tensorflow"}],"trust":{"provenance":{"is_fork":false,"github_id":99846383,"owner_type":"Organization","methodology":"github_public_v1","parent_repo":null,"near_duplicate_slugs":[]},"computed_at":"2026-08-03T00:02:03.039Z","maintenance":{"label":"Archived","score":8,"methodology":"github_public_v1","releases_90d":0,"days_since_push":4,"last_release_at":"2023-06-12T09:26:22Z"},"security_summary":{"status":"no_lockfile","scanner":null,"low_count":0,"high_count":0,"last_scan_at":"2026-07-11T23:23:13.169Z","medium_count":0,"scan_profile":"none","critical_count":0}},"capability_facts":{"scan":{"source":"repo_scan","observed_at":"2026-08-03T00:02:03.469Z"},"languages":{"value":["python"],"source":"github.language","observed_at":"2026-08-03T00:02:03.469Z"},"license_spdx":{"value":"Other","source":"github.license","observed_at":"2026-08-03T00:02:03.469Z"}},"decision_facts":{"hosting":null,"pricing":null,"requirements":null,"constraints":null,"when_to_use":["When you need to scale your training across multiple GPUs or nodes with little modification to existing scripts.","For environments where simplicity and quick setup of multi-GPU training are prioritized."],"when_not_to_use":["Avoid when extensive customization beyond core training is needed, as Horovod simplifies processes which might limit flexibility.","Not recommended if your project relies heavily on specific features not well-supported in Horovod's integration with frameworks like TensorFlow or PyTorch."],"source":"enrich:decision_facts","observed_at":"2026-07-12T12:24:48.563Z"},"constraint_facets":null,"decision_summary":[{"label":"Adopt for","value":"Simplify distributed deep learning training for TensorFlow, Keras, PyTorch, and MXNet with minimal code changes."}]}}
![\"Logo\"](\"https://user-images.githubusercontent.com/16640218/34506318-84d0c06c-efe0-11e7-8831-0425772ed8f2.png\")
![\"LF](\"https://raw.githubusercontent.com/lfai/artwork/master/lfaidata-assets/lfaidata-project-badge/graduate/color/lfaidata-project-badge-graduate-color.png\"){kind=icon}