{"id":18616722,"url":"https://github.com/sile/nasbench-rs","last_synced_at":"2025-12-12T13:58:51.703Z","repository":{"id":57643353,"uuid":"186112246","full_name":"sile/nasbench-rs","owner":"sile","description":"A Rust port of NASBench: https://github.com/google-research/nasbench","archived":false,"fork":false,"pushed_at":"2024-12-17T04:34:19.000Z","size":45,"stargazers_count":2,"open_issues_count":0,"forks_count":2,"subscribers_count":3,"default_branch":"master","last_synced_at":"2025-03-25T07:12:46.839Z","etag":null,"topics":["benchmark-datasets","neural-architecture-search","rust"],"latest_commit_sha":null,"homepage":null,"language":"Rust","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/sile.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","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":"2019-05-11T09:19:31.000Z","updated_at":"2024-12-17T04:34:23.000Z","dependencies_parsed_at":"2024-11-07T03:41:57.738Z","dependency_job_id":"5288b91d-d35a-4144-9b40-1834bc03a92b","html_url":"https://github.com/sile/nasbench-rs","commit_stats":null,"previous_names":["sile/nasbench"],"tags_count":3,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sile%2Fnasbench-rs","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sile%2Fnasbench-rs/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sile%2Fnasbench-rs/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sile%2Fnasbench-rs/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/sile","download_url":"https://codeload.github.com/sile/nasbench-rs/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":248325275,"owners_count":21084900,"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":["benchmark-datasets","neural-architecture-search","rust"],"created_at":"2024-11-07T03:37:42.963Z","updated_at":"2025-12-12T13:58:51.628Z","avatar_url":"https://github.com/sile.png","language":"Rust","funding_links":[],"categories":[],"sub_categories":[],"readme":"nasbench\n=========\n\n[![nasbench](https://img.shields.io/crates/v/nasbench.svg)](https://crates.io/crates/nasbench)\n[![Documentation](https://docs.rs/nasbench/badge.svg)](https://docs.rs/nasbench)\n[![Build Status](https://travis-ci.org/sile/nasbench-rs.svg?branch=master)](https://travis-ci.org/sile/nasbench-rs)\n[![Code Coverage](https://codecov.io/gh/sile/nasbench-rs/branch/master/graph/badge.svg)](https://codecov.io/gh/sile/nasbench/branch-rs/master)\n[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)\n\nA Rust port of [google-research/nasbench][NASBench].\n\n[NASBench]: https://github.com/google-research/nasbench\n\nMotivations\n-----------\n\nOf course, the primary purpose of this crate is to make [NASBench] dataset available in Rust.\nBesides, another aim is to reduce dataset loading time.\nBy using a compact binary data format, this crate can reduce the loading time drastically.\nFor example, on my laptop, [google-research/nasbench][NASBench] requires about 200 seconds\nfor loading the full dataset. By contrast, this crate only needs a few seconds to complete the loading.\n\nExamples\n--------\n\nFirst of all, you have to convert a NASBench dataset to this crate's format as follows:\n```console\n$ wget https://storage.googleapis.com/nasbench/nasbench_full.tfrecord\n$ nasbench convert nasbench_full.tfrecord nasbench_full.bin\n$ ls -lh\n-rw-rw-rw- 1 foo foo 328M May 12 16:47 nasbench_full.bin\n-rw-rw-rw- 1 foo foo 2.0G May 12 16:45 nasbench_full.tfrecord\n```\n\nThen, you can query the evaluation metrics associated with a model (`ops` and `adjacency`):\n```console\n$ nasbench query nasbench_full.bin \\\n                 --adjacency 0100110001000000010010000010000001000000010000000 \\\n                 --ops input conv3x3-bn-relu maxpool3x3 conv3x3-bn-relu \\\n                       conv3x3-bn-relu conv1x1-bn-relu output\nEvaluationMetrics {\n    training_time: 1769.1279296875,\n    training_accuracy: 1.0,\n    validation_accuracy: 0.9241786599159241,\n    test_accuracy: 0.9211738705635071\n}\n```\n\nRust code corresponded to the above command:\n```rust\nuse nasbench::{AdjacencyMatrix, ModelSpec, NasBench, Op};\n\n// Loads the dataset.\nlet nasbench = NasBench::new(\"nasbench_full.bin\")?;\n\n// Queries a model.\nlet ops = vec![Op::Input, Op::Conv3x3, Op::MaxPool3x3, Op::Conv3x3,\n               Op::Conv3x3, Op::Conv1x1, Op::Output];\nlet adjacency = \"0100110001000000010010000010000001000000010000000\".parse()?;\nlet model_spec = ModelSpec::new(ops, adjacency)?;\nprintln!(\"{:?}\", nasbench.models().get(\u0026model_spec));\n```\n\nLimitations\n------------\n\n[google-research/nasbench][NASBench] provides `NASBench.evaluate()` method to train and evaluate\nmodels from scratch, but this crate does not.\n\nReferences\n----------\n\n- [NAS-Bench-101: Towards Reproducible Neural Architecture Search](https://arxiv.org/abs/1902.09635)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsile%2Fnasbench-rs","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsile%2Fnasbench-rs","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsile%2Fnasbench-rs/lists"}