{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NMAIBTIFYQICL6UK3HVGILVJOE","short_pith_number":"pith:NMAIBTIF","schema_version":"1.0","canonical_sha256":"6b0080cd05c41025fa8ad9ea642ea971232c5596325c5a5af338b54a8cbfb017","source":{"kind":"arxiv","id":"2509.02480","version":1},"attestation_state":"computed","paper":{"title":"MLP-Offload: Multi-Level, Multi-Path Offloading for LLM Pre-training to Break the GPU Memory Wall","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Avinash Maurya, Bogdan Nicolae, Franck Cappello, M. Mustafa Rafique","submitted_at":"2025-09-02T16:30:49Z","abstract_excerpt":"Training LLMs larger than the aggregated memory of multiple GPUs is increasingly necessary due to the faster growth of LLM sizes compared to GPU memory. To this end, multi-tier host memory or disk offloading techniques are proposed by state of art. Despite advanced asynchronous multi-tier read/write strategies, such offloading strategies result in significant I/O overheads in the critical path of training, resulting in slower iterations. To this end, we propose MLP-Offload, a novel multi-level, multi-path offloading engine specifically designed for optimizing LLM training on resource-constrain"},"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":"2509.02480","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2025-09-02T16:30:49Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"d1876be30a22827669872bd3176e43073c865bcca4c501e80b4df4268caa21f4","abstract_canon_sha256":"fb7840a25e6ab58ac16d11b4f1c8e1a8b366ca8f8fc32327a000f78eeae49b18"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:03:39.290966Z","signature_b64":"Xk4Z85/wvyIoCMO9lDvL/X5x626HxUPoJIM8/+SNITX6IhPX/InMuXJ3Xvktv2Uy5Qgw6Vyo051ZHrRpevK9CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6b0080cd05c41025fa8ad9ea642ea971232c5596325c5a5af338b54a8cbfb017","last_reissued_at":"2026-07-05T12:03:39.290290Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:03:39.290290Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MLP-Offload: Multi-Level, Multi-Path Offloading for LLM Pre-training to Break the GPU Memory Wall","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Avinash Maurya, Bogdan Nicolae, Franck Cappello, M. Mustafa Rafique","submitted_at":"2025-09-02T16:30:49Z","abstract_excerpt":"Training LLMs larger than the aggregated memory of multiple GPUs is increasingly necessary due to the faster growth of LLM sizes compared to GPU memory. To this end, multi-tier host memory or disk offloading techniques are proposed by state of art. Despite advanced asynchronous multi-tier read/write strategies, such offloading strategies result in significant I/O overheads in the critical path of training, resulting in slower iterations. To this end, we propose MLP-Offload, a novel multi-level, multi-path offloading engine specifically designed for optimizing LLM training on resource-constrain"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.02480","kind":"arxiv","version":1},"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/2509.02480/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":"2509.02480","created_at":"2026-07-05T12:03:39.290395+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.02480v1","created_at":"2026-07-05T12:03:39.290395+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.02480","created_at":"2026-07-05T12:03:39.290395+00:00"},{"alias_kind":"pith_short_12","alias_value":"NMAIBTIFYQIC","created_at":"2026-07-05T12:03:39.290395+00:00"},{"alias_kind":"pith_short_16","alias_value":"NMAIBTIFYQICL6UK","created_at":"2026-07-05T12:03:39.290395+00:00"},{"alias_kind":"pith_short_8","alias_value":"NMAIBTIF","created_at":"2026-07-05T12:03:39.290395+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE","json":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE.json","graph_json":"https://pith.science/api/pith-number/NMAIBTIFYQICL6UK3HVGILVJOE/graph.json","events_json":"https://pith.science/api/pith-number/NMAIBTIFYQICL6UK3HVGILVJOE/events.json","paper":"https://pith.science/paper/NMAIBTIF"},"agent_actions":{"view_html":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE","download_json":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE.json","view_paper":"https://pith.science/paper/NMAIBTIF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.02480&json=true","fetch_graph":"https://pith.science/api/pith-number/NMAIBTIFYQICL6UK3HVGILVJOE/graph.json","fetch_events":"https://pith.science/api/pith-number/NMAIBTIFYQICL6UK3HVGILVJOE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE/action/storage_attestation","attest_author":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE/action/author_attestation","sign_citation":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE/action/citation_signature","submit_replication":"https://pith.science/pith/NMAIBTIFYQICL6UK3HVGILVJOE/action/replication_record"}},"created_at":"2026-07-05T12:03:39.290395+00:00","updated_at":"2026-07-05T12:03:39.290395+00:00"}