{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:HH2MM2EJGG5IOL3PCPN6EV4EHO","short_pith_number":"pith:HH2MM2EJ","schema_version":"1.0","canonical_sha256":"39f4c6688931ba872f6f13dbe257843bb1552dc160b86de035a192e75e21d357","source":{"kind":"arxiv","id":"2405.14009","version":2},"attestation_state":"computed","paper":{"title":"ReCycle: Resilient Training of Large DNNs using Pipeline Adaptation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Athinagoras Skiadopoulos, Christos Kozyrakis, Mark Zhao, Swapnil Gandhi","submitted_at":"2024-05-22T21:35:56Z","abstract_excerpt":"Training large Deep Neural Network (DNN) models requires thousands of GPUs over the course of several days or weeks. At this scale, failures are frequent and can have a big impact on training throughput. Utilizing spare GPU servers to mitigate performance loss becomes increasingly costly as model sizes grow. ReCycle is a system designed for efficient DNN training in the presence of failures, without relying on spare servers. It exploits the inherent functional redundancy in distributed training systems -- where servers across data-parallel groups store the same model parameters -- and pipeline"},"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":"2405.14009","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2024-05-22T21:35:56Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"8d1514e53fc378a3450a96900d148f889ba15327aabf28cb4ffd7106c3d68f84","abstract_canon_sha256":"931284f09d73c7e05c249c684c991456a1edd28657bf774134759e0bdfa71bad"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:11:47.336197Z","signature_b64":"YVWJNna3vU9sDBMXI13hgDdHfxIwjS6rlyk1yRg5pKop9L8Zo6mpQ7fQoxRQwPHPkveIx3iipw8yP9MoER7ODg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"39f4c6688931ba872f6f13dbe257843bb1552dc160b86de035a192e75e21d357","last_reissued_at":"2026-07-05T09:11:47.335664Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:11:47.335664Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ReCycle: Resilient Training of Large DNNs using Pipeline Adaptation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Athinagoras Skiadopoulos, Christos Kozyrakis, Mark Zhao, Swapnil Gandhi","submitted_at":"2024-05-22T21:35:56Z","abstract_excerpt":"Training large Deep Neural Network (DNN) models requires thousands of GPUs over the course of several days or weeks. At this scale, failures are frequent and can have a big impact on training throughput. Utilizing spare GPU servers to mitigate performance loss becomes increasingly costly as model sizes grow. ReCycle is a system designed for efficient DNN training in the presence of failures, without relying on spare servers. It exploits the inherent functional redundancy in distributed training systems -- where servers across data-parallel groups store the same model parameters -- and pipeline"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.14009","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/2405.14009/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":"2405.14009","created_at":"2026-07-05T09:11:47.335731+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.14009v2","created_at":"2026-07-05T09:11:47.335731+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.14009","created_at":"2026-07-05T09:11:47.335731+00:00"},{"alias_kind":"pith_short_12","alias_value":"HH2MM2EJGG5I","created_at":"2026-07-05T09:11:47.335731+00:00"},{"alias_kind":"pith_short_16","alias_value":"HH2MM2EJGG5IOL3P","created_at":"2026-07-05T09:11:47.335731+00:00"},{"alias_kind":"pith_short_8","alias_value":"HH2MM2EJ","created_at":"2026-07-05T09:11:47.335731+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01646","citing_title":"PHOENIX: Resilient LLM Training with Hot-Swapping via Zero-Overhead Checkpoint","ref_index":37,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO","json":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO.json","graph_json":"https://pith.science/api/pith-number/HH2MM2EJGG5IOL3PCPN6EV4EHO/graph.json","events_json":"https://pith.science/api/pith-number/HH2MM2EJGG5IOL3PCPN6EV4EHO/events.json","paper":"https://pith.science/paper/HH2MM2EJ"},"agent_actions":{"view_html":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO","download_json":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO.json","view_paper":"https://pith.science/paper/HH2MM2EJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.14009&json=true","fetch_graph":"https://pith.science/api/pith-number/HH2MM2EJGG5IOL3PCPN6EV4EHO/graph.json","fetch_events":"https://pith.science/api/pith-number/HH2MM2EJGG5IOL3PCPN6EV4EHO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO/action/storage_attestation","attest_author":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO/action/author_attestation","sign_citation":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO/action/citation_signature","submit_replication":"https://pith.science/pith/HH2MM2EJGG5IOL3PCPN6EV4EHO/action/replication_record"}},"created_at":"2026-07-05T09:11:47.335731+00:00","updated_at":"2026-07-05T09:11:47.335731+00:00"}