{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:ORTWBKL6HYQULAR5XLXUYNISFI","short_pith_number":"pith:ORTWBKL6","schema_version":"1.0","canonical_sha256":"746760a97e3e2145823dbaef4c35122a063bd2d09428d1b2a6db4913abb01d98","source":{"kind":"arxiv","id":"2209.13258","version":4},"attestation_state":"computed","paper":{"title":"OSDP: Optimal Sharded Data Parallel for Distributed Deep Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Bin Cui, Fangcheng Fu, Xiaonan Nie, Xupeng Miao, Youhe Jiang","submitted_at":"2022-09-27T08:55:57Z","abstract_excerpt":"Large-scale deep learning models contribute to significant performance improvements on varieties of downstream tasks. Current data and model parallelism approaches utilize model replication and partition techniques to support the distributed training of ultra-large models. However, directly deploying these systems often leads to sub-optimal training efficiency due to the complex model architectures and the strict device memory constraints. In this paper, we propose Optimal Sharded Data Parallel (OSDP), an automated parallel training system that combines the advantages from both data and model "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2209.13258","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2022-09-27T08:55:57Z","cross_cats_sorted":[],"title_canon_sha256":"8ce28bf398abbf22384ab2815a05077764c540f59bd9e0405a5d986f69b6caa1","abstract_canon_sha256":"bbb75708017e2d2f9c74d4c0b00a3e80cda6a6054ebbc5fb82188cb7fb3eb211"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:11:32.168299Z","signature_b64":"s2z8qEZXBgBvaCfMepPRUarz6VNkweTvZ1Z9ZN2nY9ghC1qOrYp0yhZ0fKfikaKajBwLubTzfdtR/3pwmQPRAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"746760a97e3e2145823dbaef4c35122a063bd2d09428d1b2a6db4913abb01d98","last_reissued_at":"2026-07-05T06:11:32.167889Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:11:32.167889Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"OSDP: Optimal Sharded Data Parallel for Distributed Deep Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Bin Cui, Fangcheng Fu, Xiaonan Nie, Xupeng Miao, Youhe Jiang","submitted_at":"2022-09-27T08:55:57Z","abstract_excerpt":"Large-scale deep learning models contribute to significant performance improvements on varieties of downstream tasks. Current data and model parallelism approaches utilize model replication and partition techniques to support the distributed training of ultra-large models. However, directly deploying these systems often leads to sub-optimal training efficiency due to the complex model architectures and the strict device memory constraints. In this paper, we propose Optimal Sharded Data Parallel (OSDP), an automated parallel training system that combines the advantages from both data and model "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2209.13258","kind":"arxiv","version":4},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2209.13258/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2209.13258","created_at":"2026-07-05T06:11:32.167948+00:00"},{"alias_kind":"arxiv_version","alias_value":"2209.13258v4","created_at":"2026-07-05T06:11:32.167948+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2209.13258","created_at":"2026-07-05T06:11:32.167948+00:00"},{"alias_kind":"pith_short_12","alias_value":"ORTWBKL6HYQU","created_at":"2026-07-05T06:11:32.167948+00:00"},{"alias_kind":"pith_short_16","alias_value":"ORTWBKL6HYQULAR5","created_at":"2026-07-05T06:11:32.167948+00:00"},{"alias_kind":"pith_short_8","alias_value":"ORTWBKL6","created_at":"2026-07-05T06:11:32.167948+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2309.10305","citing_title":"Baichuan 2: Open Large-scale Language Models","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16637","citing_title":"HexAGenT: Efficient Agentic LLM Serving via Workflow- and Heterogeneity-Aware Scheduling","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07144","citing_title":"Autopoiesis: A Self-Evolving System Paradigm for LLM Serving Under Runtime Dynamics","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07569","citing_title":"HexiSeq: Accommodating Long Context Training of LLMs over Heterogeneous Hardware","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI","json":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI.json","graph_json":"https://pith.science/api/pith-number/ORTWBKL6HYQULAR5XLXUYNISFI/graph.json","events_json":"https://pith.science/api/pith-number/ORTWBKL6HYQULAR5XLXUYNISFI/events.json","paper":"https://pith.science/paper/ORTWBKL6"},"agent_actions":{"view_html":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI","download_json":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI.json","view_paper":"https://pith.science/paper/ORTWBKL6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2209.13258&json=true","fetch_graph":"https://pith.science/api/pith-number/ORTWBKL6HYQULAR5XLXUYNISFI/graph.json","fetch_events":"https://pith.science/api/pith-number/ORTWBKL6HYQULAR5XLXUYNISFI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI/action/storage_attestation","attest_author":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI/action/author_attestation","sign_citation":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI/action/citation_signature","submit_replication":"https://pith.science/pith/ORTWBKL6HYQULAR5XLXUYNISFI/action/replication_record"}},"created_at":"2026-07-05T06:11:32.167948+00:00","updated_at":"2026-07-05T06:11:32.167948+00:00"}