{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:5V6VY46GOSN7OXHZUNOYFC2GGQ","short_pith_number":"pith:5V6VY46G","schema_version":"1.0","canonical_sha256":"ed7d5c73c6749bf75cf9a35d828b46341efd81b925e6cf40d9f72fee85965600","source":{"kind":"arxiv","id":"2010.08899","version":2},"attestation_state":"computed","paper":{"title":"Training Recommender Systems at Scale: Communication-Efficient Model and Data Parallelism","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Arun Kejariwal, Dhruv Choudhary, Kannan Ramchandran, Michael W. Mahoney, Ping Tak Peter Tang, Vipul Gupta, Xiaohan Wei, Xing Wang, Yuzhen Huang","submitted_at":"2020-10-18T01:44:42Z","abstract_excerpt":"In this paper, we consider hybrid parallelism -- a paradigm that employs both Data Parallelism (DP) and Model Parallelism (MP) -- to scale distributed training of large recommendation models. We propose a compression framework called Dynamic Communication Thresholding (DCT) for communication-efficient hybrid training. DCT filters the entities to be communicated across the network through a simple hard-thresholding function, allowing only the most relevant information to pass through. For communication efficient DP, DCT compresses the parameter gradients sent to the parameter server during mode"},"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":"2010.08899","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-18T01:44:42Z","cross_cats_sorted":["cs.DC","stat.ML"],"title_canon_sha256":"df0422b80b5944553d9478ef6d01b0470766f669f59e75b160766f0234352924","abstract_canon_sha256":"d1fbe704a750b41c447ce50a70fe17e3f63b0594926c763c31c1d23f96cf983c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:42:02.451921Z","signature_b64":"p4mrenuFyec3AjnJHu/qVfdMBn41pq+pmXkLysf+myBLm0pHAV9yl5g+mnqJ7rAuyx+WVGJ6l9nk+RKCoLdDCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ed7d5c73c6749bf75cf9a35d828b46341efd81b925e6cf40d9f72fee85965600","last_reissued_at":"2026-07-05T02:42:02.451403Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:42:02.451403Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training Recommender Systems at Scale: Communication-Efficient Model and Data Parallelism","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Arun Kejariwal, Dhruv Choudhary, Kannan Ramchandran, Michael W. Mahoney, Ping Tak Peter Tang, Vipul Gupta, Xiaohan Wei, Xing Wang, Yuzhen Huang","submitted_at":"2020-10-18T01:44:42Z","abstract_excerpt":"In this paper, we consider hybrid parallelism -- a paradigm that employs both Data Parallelism (DP) and Model Parallelism (MP) -- to scale distributed training of large recommendation models. We propose a compression framework called Dynamic Communication Thresholding (DCT) for communication-efficient hybrid training. DCT filters the entities to be communicated across the network through a simple hard-thresholding function, allowing only the most relevant information to pass through. For communication efficient DP, DCT compresses the parameter gradients sent to the parameter server during mode"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.08899","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/2010.08899/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":"2010.08899","created_at":"2026-07-05T02:42:02.451458+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.08899v2","created_at":"2026-07-05T02:42:02.451458+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.08899","created_at":"2026-07-05T02:42:02.451458+00:00"},{"alias_kind":"pith_short_12","alias_value":"5V6VY46GOSN7","created_at":"2026-07-05T02:42:02.451458+00:00"},{"alias_kind":"pith_short_16","alias_value":"5V6VY46GOSN7OXHZ","created_at":"2026-07-05T02:42:02.451458+00:00"},{"alias_kind":"pith_short_8","alias_value":"5V6VY46G","created_at":"2026-07-05T02:42:02.451458+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.05744","citing_title":"DISCD: Distributed Lossy Semantic Communication for Logical Deduction of Hypothesis","ref_index":13,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ","json":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ.json","graph_json":"https://pith.science/api/pith-number/5V6VY46GOSN7OXHZUNOYFC2GGQ/graph.json","events_json":"https://pith.science/api/pith-number/5V6VY46GOSN7OXHZUNOYFC2GGQ/events.json","paper":"https://pith.science/paper/5V6VY46G"},"agent_actions":{"view_html":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ","download_json":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ.json","view_paper":"https://pith.science/paper/5V6VY46G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.08899&json=true","fetch_graph":"https://pith.science/api/pith-number/5V6VY46GOSN7OXHZUNOYFC2GGQ/graph.json","fetch_events":"https://pith.science/api/pith-number/5V6VY46GOSN7OXHZUNOYFC2GGQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ/action/storage_attestation","attest_author":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ/action/author_attestation","sign_citation":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ/action/citation_signature","submit_replication":"https://pith.science/pith/5V6VY46GOSN7OXHZUNOYFC2GGQ/action/replication_record"}},"created_at":"2026-07-05T02:42:02.451458+00:00","updated_at":"2026-07-05T02:42:02.451458+00:00"}