{"id":18701905,"url":"https://github.com/iceychris/sps-pytorch","last_synced_at":"2026-04-25T23:33:57.145Z","repository":{"id":217920760,"uuid":"256203265","full_name":"iceychris/sps-pytorch","owner":"iceychris","description":":eight_spoked_asterisk: An unofficial implementation of Stochastic Polyak Step-size in PyTorch","archived":false,"fork":false,"pushed_at":"2020-04-16T14:39:30.000Z","size":173,"stargazers_count":2,"open_issues_count":0,"forks_count":1,"subscribers_count":1,"default_branch":"master","last_synced_at":"2025-10-26T00:30:16.528Z","etag":null,"topics":["optimizer","pytorch","reproduction","sota","sps"],"latest_commit_sha":null,"homepage":"https://arxiv.org/abs/2002.10542","language":"Python","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/iceychris.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}},"created_at":"2020-04-16T12:06:12.000Z","updated_at":"2023-09-08T18:06:02.000Z","dependencies_parsed_at":null,"dependency_job_id":"42e1910a-11a9-4c3e-89eb-30c1dab5201b","html_url":"https://github.com/iceychris/sps-pytorch","commit_stats":{"total_commits":7,"total_committers":1,"mean_commits":7.0,"dds":0.0,"last_synced_commit":"7fda91642bb8604dce5ca0a12fdd8f542ab7829f"},"previous_names":["iceychris/sps-pytorch"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/iceychris/sps-pytorch","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/iceychris%2Fsps-pytorch","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/iceychris%2Fsps-pytorch/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/iceychris%2Fsps-pytorch/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/iceychris%2Fsps-pytorch/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/iceychris","download_url":"https://codeload.github.com/iceychris/sps-pytorch/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/iceychris%2Fsps-pytorch/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":32280974,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-25T18:29:39.964Z","status":"ssl_error","status_checked_at":"2026-04-25T18:29:32.149Z","response_time":59,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"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":["optimizer","pytorch","reproduction","sota","sps"],"created_at":"2024-11-07T11:42:59.460Z","updated_at":"2026-04-25T23:33:57.131Z","avatar_url":"https://github.com/iceychris.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# sps-pytorch\n\n[Please find the official implementation of the authors here](https://github.com/IssamLaradji/sps).\n\nThis repo aims to reproduce and extend the `SPS` optimizer by `Nicolas Loizou`, `Sharan Vaswani`, `Issam Laradji` and `Simon Lacoste-Julien`\npresented in their paper titled [`Stochastic Polyak Step-size for SGD: An Adaptive Learning Rate for Fast Convergence`](https://arxiv.org/abs/2002.10542).\n\n\n## Results\n\n| RNN-T |\n|:-----:|\n|![](./images/rnn-t_loss_train.png)|\n|![](./images/rnn-t_loss_valid.png)|\n|![step size sps](./images/rnn-t_step_size_sps.png)|\n|![step size over9000](./images/rnn-t_step_size_over9000.png)|\n\n\n## Run\n\n- clone and patch fastai\n\n```bash\n# clone\ngit clone https://github.com/fastai/fastai\ncd fastai\ngit checkout 1.0.60\npip install -e \".[dev]\"\n\n# patch\ngit apply fastai.patch\n```\n\n- `python train.py --opt sps --epochs 5`\n\n\n## Resources\n\n- [official Implementation of `SPS`](https://github.com/IssamLaradji/sps)\n\n- [mgrankin/over9000](https://github.com/mgrankin/over9000) (code snippets)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ficeychris%2Fsps-pytorch","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ficeychris%2Fsps-pytorch","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ficeychris%2Fsps-pytorch/lists"}