{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LMCEGUPXGPLH6N7V2K37FEMYT6","short_pith_number":"pith:LMCEGUPX","schema_version":"1.0","canonical_sha256":"5b044351f733d67f37f5d2b7f291989fb69cb3d92c587fc3682d08b778750d0a","source":{"kind":"arxiv","id":"2406.17145","version":2},"attestation_state":"computed","paper":{"title":"GraphPipe: Improving Performance and Scalability of DNN Training with Graph Pipeline Parallelism","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Byungsoo Jeon, Colin Unger, Daiyaan Arfeen, Gregory R. Ganger, Mengdi Wu, Mohammad Alizadeh, Neeraj Aggarwal, Peiyuan Liao, Shiyi Cao, Sunghyun Kim, Sunghyun Park, Tianqi Chen, Xupeng Miao, Zhihao Jia","submitted_at":"2024-06-24T21:32:51Z","abstract_excerpt":"Deep neural networks (DNNs) continue to grow rapidly in size, making them infeasible to train on a single device. Pipeline parallelism is commonly used in existing DNN systems to support large-scale DNN training by partitioning a DNN into multiple stages, which concurrently perform DNN training for different micro-batches in a pipeline fashion. However, existing pipeline-parallel approaches only consider sequential pipeline stages and thus ignore the topology of a DNN, resulting in missed model-parallel opportunities. This paper presents graph pipeline parallelism (GPP), a new pipeline-paralle"},"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":"2406.17145","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2024-06-24T21:32:51Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"25a6548484a5e7601b8d70df0dbfdf082f4c615fe435ffbbc467923aa69cdef0","abstract_canon_sha256":"4b982c2292edc07230f212dc67ca28f641888daddaa208a6f0f852738d1cb29d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:27:02.530014Z","signature_b64":"Q11ny8Rc8ycvML+US7e6cqHsfZ0KX3pdlMgOfy419G0qbXKaSGwaGrt5q+XVt+4PGbwh0qylL/+uTniBzC1EDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5b044351f733d67f37f5d2b7f291989fb69cb3d92c587fc3682d08b778750d0a","last_reissued_at":"2026-07-05T09:27:02.529472Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:27:02.529472Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GraphPipe: Improving Performance and Scalability of DNN Training with Graph Pipeline Parallelism","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Byungsoo Jeon, Colin Unger, Daiyaan Arfeen, Gregory R. Ganger, Mengdi Wu, Mohammad Alizadeh, Neeraj Aggarwal, Peiyuan Liao, Shiyi Cao, Sunghyun Kim, Sunghyun Park, Tianqi Chen, Xupeng Miao, Zhihao Jia","submitted_at":"2024-06-24T21:32:51Z","abstract_excerpt":"Deep neural networks (DNNs) continue to grow rapidly in size, making them infeasible to train on a single device. Pipeline parallelism is commonly used in existing DNN systems to support large-scale DNN training by partitioning a DNN into multiple stages, which concurrently perform DNN training for different micro-batches in a pipeline fashion. However, existing pipeline-parallel approaches only consider sequential pipeline stages and thus ignore the topology of a DNN, resulting in missed model-parallel opportunities. This paper presents graph pipeline parallelism (GPP), a new pipeline-paralle"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.17145","kind":"arxiv","version":2},"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/2406.17145/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":"2406.17145","created_at":"2026-07-05T09:27:02.529542+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.17145v2","created_at":"2026-07-05T09:27:02.529542+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.17145","created_at":"2026-07-05T09:27:02.529542+00:00"},{"alias_kind":"pith_short_12","alias_value":"LMCEGUPXGPLH","created_at":"2026-07-05T09:27:02.529542+00:00"},{"alias_kind":"pith_short_16","alias_value":"LMCEGUPXGPLH6N7V","created_at":"2026-07-05T09:27:02.529542+00:00"},{"alias_kind":"pith_short_8","alias_value":"LMCEGUPX","created_at":"2026-07-05T09:27:02.529542+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19163","citing_title":"Pulse: Training Acceleration for Large Diffusion Models with Automatic Pipeline Parallelism","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27678","citing_title":"Heterogeneous Parallelism for Multimodal Large Language Model Training","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08962","citing_title":"MegaScale-Omni: A Hyper-Scale, Workload-Resilient System for MultiModal LLM Training in Production","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6","json":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6.json","graph_json":"https://pith.science/api/pith-number/LMCEGUPXGPLH6N7V2K37FEMYT6/graph.json","events_json":"https://pith.science/api/pith-number/LMCEGUPXGPLH6N7V2K37FEMYT6/events.json","paper":"https://pith.science/paper/LMCEGUPX"},"agent_actions":{"view_html":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6","download_json":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6.json","view_paper":"https://pith.science/paper/LMCEGUPX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.17145&json=true","fetch_graph":"https://pith.science/api/pith-number/LMCEGUPXGPLH6N7V2K37FEMYT6/graph.json","fetch_events":"https://pith.science/api/pith-number/LMCEGUPXGPLH6N7V2K37FEMYT6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6/action/storage_attestation","attest_author":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6/action/author_attestation","sign_citation":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6/action/citation_signature","submit_replication":"https://pith.science/pith/LMCEGUPXGPLH6N7V2K37FEMYT6/action/replication_record"}},"created_at":"2026-07-05T09:27:02.529542+00:00","updated_at":"2026-07-05T09:27:02.529542+00:00"}