{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:5MNN3ESPUBUZ7RMNTFRAQPY7ZU","short_pith_number":"pith:5MNN3ESP","schema_version":"1.0","canonical_sha256":"eb1add924fa0699fc58d9962083f1fcd2dabc668e246f73ee7cebbccfe16132b","source":{"kind":"arxiv","id":"1711.04325","version":1},"attestation_state":"computed","paper":{"title":"Extremely Large Minibatch SGD: Training ResNet-50 on ImageNet in 15 Minutes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.DC","authors_text":"Keisuke Fukuda, Shuji Suzuki, Takuya Akiba","submitted_at":"2017-11-12T17:36:46Z","abstract_excerpt":"We demonstrate that training ResNet-50 on ImageNet for 90 epochs can be achieved in 15 minutes with 1024 Tesla P100 GPUs. This was made possible by using a large minibatch size of 32k. To maintain accuracy with this large minibatch size, we employed several techniques such as RMSprop warm-up, batch normalization without moving averages, and a slow-start learning rate schedule. This paper also describes the details of the hardware and software of the system used to achieve the above performance."},"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":"1711.04325","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2017-11-12T17:36:46Z","cross_cats_sorted":["cs.CV","cs.LG"],"title_canon_sha256":"d43e844c812acc66ef0e5a7b4f36929acb0bc4620e78d325733d7d67bba69109","abstract_canon_sha256":"6012d0608f9b15b571e762c223a652bff5846e181e177e390e6f3babfb629de8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:30:39.406788Z","signature_b64":"Cw7QtgAi3RibRyu6biIko+9bIBjZyAxoZ3u74opffnGyA7jz+QqJIXxY852KUcQRe9zIPoU1AL76RC99ZqbVBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eb1add924fa0699fc58d9962083f1fcd2dabc668e246f73ee7cebbccfe16132b","last_reissued_at":"2026-05-18T00:30:39.406125Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:30:39.406125Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Extremely Large Minibatch SGD: Training ResNet-50 on ImageNet in 15 Minutes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.DC","authors_text":"Keisuke Fukuda, Shuji Suzuki, Takuya Akiba","submitted_at":"2017-11-12T17:36:46Z","abstract_excerpt":"We demonstrate that training ResNet-50 on ImageNet for 90 epochs can be achieved in 15 minutes with 1024 Tesla P100 GPUs. This was made possible by using a large minibatch size of 32k. To maintain accuracy with this large minibatch size, we employed several techniques such as RMSprop warm-up, batch normalization without moving averages, and a slow-start learning rate schedule. This paper also describes the details of the hardware and software of the system used to achieve the above performance."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1711.04325","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":""},"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":"1711.04325","created_at":"2026-05-18T00:30:39.406236+00:00"},{"alias_kind":"arxiv_version","alias_value":"1711.04325v1","created_at":"2026-05-18T00:30:39.406236+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1711.04325","created_at":"2026-05-18T00:30:39.406236+00:00"},{"alias_kind":"pith_short_12","alias_value":"5MNN3ESPUBUZ","created_at":"2026-05-18T12:31:00.734936+00:00"},{"alias_kind":"pith_short_16","alias_value":"5MNN3ESPUBUZ7RMN","created_at":"2026-05-18T12:31:00.734936+00:00"},{"alias_kind":"pith_short_8","alias_value":"5MNN3ESP","created_at":"2026-05-18T12:31:00.734936+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"1906.10822","citing_title":"Gradient Noise Convolution (GNC): Smoothing Loss Function for Distributed Large-Batch SGD","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"1906.11786","citing_title":"Fast Training of Sparse Graph Neural Networks on Dense Hardware","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"1904.00962","citing_title":"Large Batch Optimization for Deep Learning: Training BERT in 76 minutes","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2102.01293","citing_title":"Scaling Laws for Transfer","ref_index":125,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02473","citing_title":"Analyzing Reverse Address Translation Overheads in Multi-GPU Scale-Up Pods","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2112.00861","citing_title":"A General Language Assistant as a Laboratory for Alignment","ref_index":167,"is_internal_anchor":false},{"citing_arxiv_id":"2207.05221","citing_title":"Language Models (Mostly) Know What They Know","ref_index":244,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU","json":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU.json","graph_json":"https://pith.science/api/pith-number/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/graph.json","events_json":"https://pith.science/api/pith-number/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/events.json","paper":"https://pith.science/paper/5MNN3ESP"},"agent_actions":{"view_html":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU","download_json":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU.json","view_paper":"https://pith.science/paper/5MNN3ESP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1711.04325&json=true","fetch_graph":"https://pith.science/api/pith-number/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/graph.json","fetch_events":"https://pith.science/api/pith-number/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/action/storage_attestation","attest_author":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/action/author_attestation","sign_citation":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/action/citation_signature","submit_replication":"https://pith.science/pith/5MNN3ESPUBUZ7RMNTFRAQPY7ZU/action/replication_record"}},"created_at":"2026-05-18T00:30:39.406236+00:00","updated_at":"2026-05-18T00:30:39.406236+00:00"}