{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:BIMLI3THH5PSHA56CBSKDVHQKR","short_pith_number":"pith:BIMLI3TH","schema_version":"1.0","canonical_sha256":"0a18b46e673f5f2383be1064a1d4f054422f4d1a5fb0eb7f6702c16c46100be1","source":{"kind":"arxiv","id":"1709.05011","version":10},"attestation_state":"computed","paper":{"title":"ImageNet Training in Minutes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Cho-Jui Hsieh, James Demmel, Kurt Keutzer, Yang You, Zhao Zhang","submitted_at":"2017-09-14T23:59:52Z","abstract_excerpt":"Finishing 90-epoch ImageNet-1k training with ResNet-50 on a NVIDIA M40 GPU takes 14 days. This training requires 10^18 single precision operations in total. On the other hand, the world's current fastest supercomputer can finish 2 * 10^17 single precision operations per second (Dongarra et al 2017, https://www.top500.org/lists/2017/06/). If we can make full use of the supercomputer for DNN training, we should be able to finish the 90-epoch ResNet-50 training in one minute. However, the current bottleneck for fast DNN training is in the algorithm level. Specifically, the current batch size (e.g"},"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":"1709.05011","kind":"arxiv","version":10},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2017-09-14T23:59:52Z","cross_cats_sorted":[],"title_canon_sha256":"8a38f774769d04b3f7782d962f6c4abee564a703e71f17b09834e6f3ef670224","abstract_canon_sha256":"107c47691db0c35a6fe6528b37501cc780fbd67483cc4cdf318edaaf83dc04c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:24:42.451257Z","signature_b64":"tJdeLMyBSTioBaHcmcp/zoL1CXd7bBJUujexofz+/Yfqj34O5Qr94XYpxmLuD4l5B9Qk4YnN8zvODcOJWUk4Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0a18b46e673f5f2383be1064a1d4f054422f4d1a5fb0eb7f6702c16c46100be1","last_reissued_at":"2026-05-18T00:24:42.450656Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:24:42.450656Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ImageNet Training in Minutes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Cho-Jui Hsieh, James Demmel, Kurt Keutzer, Yang You, Zhao Zhang","submitted_at":"2017-09-14T23:59:52Z","abstract_excerpt":"Finishing 90-epoch ImageNet-1k training with ResNet-50 on a NVIDIA M40 GPU takes 14 days. This training requires 10^18 single precision operations in total. On the other hand, the world's current fastest supercomputer can finish 2 * 10^17 single precision operations per second (Dongarra et al 2017, https://www.top500.org/lists/2017/06/). If we can make full use of the supercomputer for DNN training, we should be able to finish the 90-epoch ResNet-50 training in one minute. However, the current bottleneck for fast DNN training is in the algorithm level. Specifically, the current batch size (e.g"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1709.05011","kind":"arxiv","version":10},"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":"1709.05011","created_at":"2026-05-18T00:24:42.450726+00:00"},{"alias_kind":"arxiv_version","alias_value":"1709.05011v10","created_at":"2026-05-18T00:24:42.450726+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1709.05011","created_at":"2026-05-18T00:24:42.450726+00:00"},{"alias_kind":"pith_short_12","alias_value":"BIMLI3THH5PS","created_at":"2026-05-18T12:31:08.081275+00:00"},{"alias_kind":"pith_short_16","alias_value":"BIMLI3THH5PSHA56","created_at":"2026-05-18T12:31:08.081275+00:00"},{"alias_kind":"pith_short_8","alias_value":"BIMLI3TH","created_at":"2026-05-18T12:31:08.081275+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"1906.10822","citing_title":"Gradient Noise Convolution (GNC): Smoothing Loss Function for Distributed Large-Batch SGD","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2508.21613","citing_title":"Chameleon: Adaptive Fault Tolerance for Distributed Training via Real-time Policy Selection","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2102.01293","citing_title":"Scaling Laws for Transfer","ref_index":140,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11111","citing_title":"ShardTensor: Domain Parallelism for Scientific Machine Learning","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2112.00861","citing_title":"A General Language Assistant as a Laboratory for Alignment","ref_index":182,"is_internal_anchor":false},{"citing_arxiv_id":"2207.05221","citing_title":"Language Models (Mostly) Know What They Know","ref_index":259,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR","json":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR.json","graph_json":"https://pith.science/api/pith-number/BIMLI3THH5PSHA56CBSKDVHQKR/graph.json","events_json":"https://pith.science/api/pith-number/BIMLI3THH5PSHA56CBSKDVHQKR/events.json","paper":"https://pith.science/paper/BIMLI3TH"},"agent_actions":{"view_html":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR","download_json":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR.json","view_paper":"https://pith.science/paper/BIMLI3TH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1709.05011&json=true","fetch_graph":"https://pith.science/api/pith-number/BIMLI3THH5PSHA56CBSKDVHQKR/graph.json","fetch_events":"https://pith.science/api/pith-number/BIMLI3THH5PSHA56CBSKDVHQKR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR/action/storage_attestation","attest_author":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR/action/author_attestation","sign_citation":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR/action/citation_signature","submit_replication":"https://pith.science/pith/BIMLI3THH5PSHA56CBSKDVHQKR/action/replication_record"}},"created_at":"2026-05-18T00:24:42.450726+00:00","updated_at":"2026-05-18T00:24:42.450726+00:00"}