{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:JOS4IE6KUZH55WH7MIHERJ3ZL7","short_pith_number":"pith:JOS4IE6K","schema_version":"1.0","canonical_sha256":"4ba5c413caa64fded8ff620e48a7795fed8ad4a51e14fe49c524766f7e6ccb9e","source":{"kind":"arxiv","id":"2208.05321","version":1},"attestation_state":"computed","paper":{"title":"A Frequency-aware Software Cache for Large Recommendation System Embeddings","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.DC","cs.LG"],"primary_cat":"cs.IR","authors_text":"Geng Zhang, Jiarui Fang, Jiatong Han, Jin Liu, Shenggui Li, Yang You, Yongbin Li, Zhengda Bian","submitted_at":"2022-08-08T12:08:05Z","abstract_excerpt":"Deep learning recommendation models (DLRMs) have been widely applied in Internet companies. The embedding tables of DLRMs are too large to fit on GPU memory entirely. We propose a GPU-based software cache approaches to dynamically manage the embedding table in the CPU and GPU memory space by leveraging the id's frequency statistics of the target dataset. Our proposed software cache is efficient in training entire DLRMs on GPU in a synchronized update manner. It is also scaled to multiple GPUs in combination with the widely used hybrid parallel training approaches. Evaluating our prototype syst"},"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":"2208.05321","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.IR","submitted_at":"2022-08-08T12:08:05Z","cross_cats_sorted":["cs.AI","cs.DC","cs.LG"],"title_canon_sha256":"ab94b7c4dd319074cb56a1f23304279c85824fcae557c2ff7d6f9eb70d27d448","abstract_canon_sha256":"92ac121773f811296471d63a2196530d6fc4fe1daaffce11d0cb15910c476efb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:47:34.822451Z","signature_b64":"hQGv2MLIIdHLQed5Np43jlyIsn46NdY9kicNxvpWRHB/5lASugYsbVrkRInN6ME/0Fg+z7uAL16yKEOb43W9CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4ba5c413caa64fded8ff620e48a7795fed8ad4a51e14fe49c524766f7e6ccb9e","last_reissued_at":"2026-07-05T04:47:34.821981Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:47:34.821981Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Frequency-aware Software Cache for Large Recommendation System Embeddings","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.DC","cs.LG"],"primary_cat":"cs.IR","authors_text":"Geng Zhang, Jiarui Fang, Jiatong Han, Jin Liu, Shenggui Li, Yang You, Yongbin Li, Zhengda Bian","submitted_at":"2022-08-08T12:08:05Z","abstract_excerpt":"Deep learning recommendation models (DLRMs) have been widely applied in Internet companies. The embedding tables of DLRMs are too large to fit on GPU memory entirely. We propose a GPU-based software cache approaches to dynamically manage the embedding table in the CPU and GPU memory space by leveraging the id's frequency statistics of the target dataset. Our proposed software cache is efficient in training entire DLRMs on GPU in a synchronized update manner. It is also scaled to multiple GPUs in combination with the widely used hybrid parallel training approaches. Evaluating our prototype syst"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2208.05321","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/2208.05321/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":"2208.05321","created_at":"2026-07-05T04:47:34.822034+00:00"},{"alias_kind":"arxiv_version","alias_value":"2208.05321v1","created_at":"2026-07-05T04:47:34.822034+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2208.05321","created_at":"2026-07-05T04:47:34.822034+00:00"},{"alias_kind":"pith_short_12","alias_value":"JOS4IE6KUZH5","created_at":"2026-07-05T04:47:34.822034+00:00"},{"alias_kind":"pith_short_16","alias_value":"JOS4IE6KUZH55WH7","created_at":"2026-07-05T04:47:34.822034+00:00"},{"alias_kind":"pith_short_8","alias_value":"JOS4IE6K","created_at":"2026-07-05T04:47:34.822034+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.30728","citing_title":"Reducing the GPU Memory Bottleneck with Lossless Compression for ML -- Extended","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7","json":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7.json","graph_json":"https://pith.science/api/pith-number/JOS4IE6KUZH55WH7MIHERJ3ZL7/graph.json","events_json":"https://pith.science/api/pith-number/JOS4IE6KUZH55WH7MIHERJ3ZL7/events.json","paper":"https://pith.science/paper/JOS4IE6K"},"agent_actions":{"view_html":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7","download_json":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7.json","view_paper":"https://pith.science/paper/JOS4IE6K","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2208.05321&json=true","fetch_graph":"https://pith.science/api/pith-number/JOS4IE6KUZH55WH7MIHERJ3ZL7/graph.json","fetch_events":"https://pith.science/api/pith-number/JOS4IE6KUZH55WH7MIHERJ3ZL7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7/action/storage_attestation","attest_author":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7/action/author_attestation","sign_citation":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7/action/citation_signature","submit_replication":"https://pith.science/pith/JOS4IE6KUZH55WH7MIHERJ3ZL7/action/replication_record"}},"created_at":"2026-07-05T04:47:34.822034+00:00","updated_at":"2026-07-05T04:47:34.822034+00:00"}