{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:GDMKE6GDP7FN2JJP5QXZBCVMRU","short_pith_number":"pith:GDMKE6GD","schema_version":"1.0","canonical_sha256":"30d8a278c37fcadd252fec2f908aac8d116f8ebe5b973601da2d4e3ba4a523f0","source":{"kind":"arxiv","id":"2006.08517","version":1},"attestation_state":"computed","paper":{"title":"The Limit of the Batch Size","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.DC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Cho-Jui Hsieh, Huan Zhang, James Demmel, Yang You, Yuhui Wang, Zhao Zhang","submitted_at":"2020-06-15T16:18:05Z","abstract_excerpt":"Large-batch training is an efficient approach for current distributed deep learning systems. It has enabled researchers to reduce the ImageNet/ResNet-50 training from 29 hours to around 1 minute. In this paper, we focus on studying the limit of the batch size. We think it may provide a guidance to AI supercomputer and algorithm designers. We provide detailed numerical optimization instructions for step-by-step comparison. Moreover, it is important to understand the generalization and optimization performance of huge batch training. Hoffer et al. introduced \"ultra-slow diffusion\" theory to larg"},"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":"2006.08517","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-06-15T16:18:05Z","cross_cats_sorted":["cs.CV","cs.DC","stat.ML"],"title_canon_sha256":"ac2474bdc60a30e08718c87c4c1295ef50d46e4162730cad3a1ed1a613230869","abstract_canon_sha256":"6489cf17e874b5d2d533b73c5f43a41ea11647ff49108151d13bb607eddf325f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:10:22.272573Z","signature_b64":"TWhAC3kQlfHNSoRD91hLN5VQrm1q/ki/dSQjQVehoARDK/dpnRInW0pbZgqzt2xUi3UlN3qmKey1adlTqB9QDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"30d8a278c37fcadd252fec2f908aac8d116f8ebe5b973601da2d4e3ba4a523f0","last_reissued_at":"2026-07-05T01:10:22.272211Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:10:22.272211Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Limit of the Batch Size","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.DC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Cho-Jui Hsieh, Huan Zhang, James Demmel, Yang You, Yuhui Wang, Zhao Zhang","submitted_at":"2020-06-15T16:18:05Z","abstract_excerpt":"Large-batch training is an efficient approach for current distributed deep learning systems. It has enabled researchers to reduce the ImageNet/ResNet-50 training from 29 hours to around 1 minute. In this paper, we focus on studying the limit of the batch size. We think it may provide a guidance to AI supercomputer and algorithm designers. We provide detailed numerical optimization instructions for step-by-step comparison. Moreover, it is important to understand the generalization and optimization performance of huge batch training. Hoffer et al. introduced \"ultra-slow diffusion\" theory to larg"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2006.08517","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/2006.08517/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":"2006.08517","created_at":"2026-07-05T01:10:22.272272+00:00"},{"alias_kind":"arxiv_version","alias_value":"2006.08517v1","created_at":"2026-07-05T01:10:22.272272+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2006.08517","created_at":"2026-07-05T01:10:22.272272+00:00"},{"alias_kind":"pith_short_12","alias_value":"GDMKE6GDP7FN","created_at":"2026-07-05T01:10:22.272272+00:00"},{"alias_kind":"pith_short_16","alias_value":"GDMKE6GDP7FN2JJP","created_at":"2026-07-05T01:10:22.272272+00:00"},{"alias_kind":"pith_short_8","alias_value":"GDMKE6GD","created_at":"2026-07-05T01:10:22.272272+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.20511","citing_title":"Collaborative Batch Size Optimization for Federated Learning","ref_index":5,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU","json":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU.json","graph_json":"https://pith.science/api/pith-number/GDMKE6GDP7FN2JJP5QXZBCVMRU/graph.json","events_json":"https://pith.science/api/pith-number/GDMKE6GDP7FN2JJP5QXZBCVMRU/events.json","paper":"https://pith.science/paper/GDMKE6GD"},"agent_actions":{"view_html":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU","download_json":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU.json","view_paper":"https://pith.science/paper/GDMKE6GD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2006.08517&json=true","fetch_graph":"https://pith.science/api/pith-number/GDMKE6GDP7FN2JJP5QXZBCVMRU/graph.json","fetch_events":"https://pith.science/api/pith-number/GDMKE6GDP7FN2JJP5QXZBCVMRU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU/action/storage_attestation","attest_author":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU/action/author_attestation","sign_citation":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU/action/citation_signature","submit_replication":"https://pith.science/pith/GDMKE6GDP7FN2JJP5QXZBCVMRU/action/replication_record"}},"created_at":"2026-07-05T01:10:22.272272+00:00","updated_at":"2026-07-05T01:10:22.272272+00:00"}