{"id":15685332,"url":"https://github.com/vfdev-5/cifar10-faster","last_synced_at":"2025-05-07T16:55:25.574Z","repository":{"id":149000698,"uuid":"170711484","full_name":"vfdev-5/cifar10-faster","owner":"vfdev-5","description":"[WIP] Demonstration of training a small ResNet on CIFAR10 to 94% test accuracy in less 20 epochs","archived":false,"fork":false,"pushed_at":"2019-02-20T16:20:53.000Z","size":46,"stargazers_count":9,"open_issues_count":0,"forks_count":1,"subscribers_count":3,"default_branch":"master","last_synced_at":"2025-03-31T12:05:02.357Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","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/vfdev-5.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":"2019-02-14T15:15:51.000Z","updated_at":"2022-04-19T14:18:35.000Z","dependencies_parsed_at":"2023-04-30T11:36:57.099Z","dependency_job_id":null,"html_url":"https://github.com/vfdev-5/cifar10-faster","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/vfdev-5%2Fcifar10-faster","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/vfdev-5%2Fcifar10-faster/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/vfdev-5%2Fcifar10-faster/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/vfdev-5%2Fcifar10-faster/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/vfdev-5","download_url":"https://codeload.github.com/vfdev-5/cifar10-faster/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":252921953,"owners_count":21825632,"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":[],"created_at":"2024-10-03T17:24:52.997Z","updated_at":"2025-05-07T16:55:25.536Z","avatar_url":"https://github.com/vfdev-5.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Faster training on CIFAR10\n\nDemonstration of training a small ResNet on CIFAR10 to 94% test accuracy in the minimal possible time and under 20 epochs.\n\nThis code is inspired by [cifar10-fast repository](https://github.com/davidcpage/cifar10-fast) and some of\nthe current code is adapted from the repository. In his [blog articles](https://www.myrtle.ai/2018/09/24/how_to_train_your_resnet/)\nDavid Page (@davidcpage) explains the choice of the model and the way to optimize the dataflow.\n\n## Project on Polyaxon platform\n\n### Create project\n```\npolyaxon project create --name=cifar10-faster --description=\"Faster training on CIFAR10\"\n```\n\n### Initialize project \n```\npolyaxon init cifar10-faster\n```\n\n## Reproduce original result\n\nTrain `fast-resnet` during 24 epochs using cutout data augmentation, SGD optimizer, piecewise linear scheduling:\n\n```\npolyaxon run -u -f plx_configs/fastresnet/xp_original_training.yaml --name=xp_original_training --tags=original\n```\n\n## Check other configurations around `fast-resnet`\n\n### [Mixup](https://arxiv.org/abs/1710.09412)\n\nWe remove cutout data augmentation and uses mixup technics:\n```\npolyaxon run -u -f plx_configs/fastresnet/xp_training_mixup.yaml --name=xp_training_mixup --tags=original,mixup\n```\n\n### [AdamW](https://arxiv.org/pdf/1711.05101.pdf)\n\nWe uses decoupled weight decay Adam optimizer instead of SGD\n```\npolyaxon run -u -f plx_configs/fastresnet/xp_training_adamw.yaml --name=xp_training_adamw--tags=original,adamw\n```\n\n## Run hyperparameter tuning\n\n```\npolyaxon run -u -f plx_configs/fastresnet/gp_hp_bo_training.yaml --name=gp_hp_bo_training --tags=lt_20\n``` \nor on WRN model\n```\npolyaxon run -u -f plx_configs/wrn/gp_hp_bo_training_wrn.yaml --name=gp_hp_bo_training_wrn --tags=lt_20,wrn\n```\n\n\n\n## Experiments on Google Colab\n\nhttps://colab.research.google.com/drive/1W1_WEtatzyn32aPSrp4t5n66PuHQW6W8\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fvfdev-5%2Fcifar10-faster","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fvfdev-5%2Fcifar10-faster","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fvfdev-5%2Fcifar10-faster/lists"}