{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:3D6EGCK4PR56BZBATZUU6P7PPE","short_pith_number":"pith:3D6EGCK4","schema_version":"1.0","canonical_sha256":"d8fc43095c7c7be0e4209e694f3fef790fb53902d3d4c2b4689090504da72647","source":{"kind":"arxiv","id":"1806.00187","version":3},"attestation_state":"computed","paper":{"title":"Scaling Neural Machine Translation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"David Grangier, Michael Auli, Myle Ott, Sergey Edunov","submitted_at":"2018-06-01T04:33:16Z","abstract_excerpt":"Sequence to sequence learning models still require several days to reach state of the art performance on large benchmark datasets using a single machine. This paper shows that reduced precision and large batch training can speedup training by nearly 5x on a single 8-GPU machine with careful tuning and implementation. On WMT'14 English-German translation, we match the accuracy of Vaswani et al. (2017) in under 5 hours when training on 8 GPUs and we obtain a new state of the art of 29.3 BLEU after training for 85 minutes on 128 GPUs. We further improve these results to 29.8 BLEU by training on t"},"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":"1806.00187","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2018-06-01T04:33:16Z","cross_cats_sorted":[],"title_canon_sha256":"43b1c1dfb0b34120f467e4f9966fb93755d56ed8ddef740bfad296376af9ec47","abstract_canon_sha256":"c28a5cfb0236bd53d83766bf7c954567bae4f7cb2c6434d54eaaa0031e2fa9b8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:06:28.623667Z","signature_b64":"Q84u4Ixii/nJNndX/Fmj2+0p9Yaz1txhDpg7yzIKzXw+5AC7VIwZ/vEY/BOzp8D3CGey2rktQdcjGqckdbc3Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d8fc43095c7c7be0e4209e694f3fef790fb53902d3d4c2b4689090504da72647","last_reissued_at":"2026-05-18T00:06:28.622995Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:06:28.622995Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Neural Machine Translation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"David Grangier, Michael Auli, Myle Ott, Sergey Edunov","submitted_at":"2018-06-01T04:33:16Z","abstract_excerpt":"Sequence to sequence learning models still require several days to reach state of the art performance on large benchmark datasets using a single machine. This paper shows that reduced precision and large batch training can speedup training by nearly 5x on a single 8-GPU machine with careful tuning and implementation. On WMT'14 English-German translation, we match the accuracy of Vaswani et al. (2017) in under 5 hours when training on 8 GPUs and we obtain a new state of the art of 29.3 BLEU after training for 85 minutes on 128 GPUs. We further improve these results to 29.8 BLEU by training on t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1806.00187","kind":"arxiv","version":3},"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":"1806.00187","created_at":"2026-05-18T00:06:28.623100+00:00"},{"alias_kind":"arxiv_version","alias_value":"1806.00187v3","created_at":"2026-05-18T00:06:28.623100+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1806.00187","created_at":"2026-05-18T00:06:28.623100+00:00"},{"alias_kind":"pith_short_12","alias_value":"3D6EGCK4PR56","created_at":"2026-05-18T12:32:02.567920+00:00"},{"alias_kind":"pith_short_16","alias_value":"3D6EGCK4PR56BZBA","created_at":"2026-05-18T12:32:02.567920+00:00"},{"alias_kind":"pith_short_8","alias_value":"3D6EGCK4","created_at":"2026-05-18T12:32:02.567920+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.20912","citing_title":"Enhancing Scientific Discourse: Machine Translation for the Scientific Domain","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2102.01293","citing_title":"Scaling Laws for Transfer","ref_index":124,"is_internal_anchor":true},{"citing_arxiv_id":"2208.07339","citing_title":"LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale","ref_index":152,"is_internal_anchor":false},{"citing_arxiv_id":"2112.00861","citing_title":"A General Language Assistant as a Laboratory for Alignment","ref_index":166,"is_internal_anchor":false},{"citing_arxiv_id":"2207.05221","citing_title":"Language Models (Mostly) Know What They Know","ref_index":243,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE","json":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE.json","graph_json":"https://pith.science/api/pith-number/3D6EGCK4PR56BZBATZUU6P7PPE/graph.json","events_json":"https://pith.science/api/pith-number/3D6EGCK4PR56BZBATZUU6P7PPE/events.json","paper":"https://pith.science/paper/3D6EGCK4"},"agent_actions":{"view_html":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE","download_json":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE.json","view_paper":"https://pith.science/paper/3D6EGCK4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1806.00187&json=true","fetch_graph":"https://pith.science/api/pith-number/3D6EGCK4PR56BZBATZUU6P7PPE/graph.json","fetch_events":"https://pith.science/api/pith-number/3D6EGCK4PR56BZBATZUU6P7PPE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE/action/storage_attestation","attest_author":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE/action/author_attestation","sign_citation":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE/action/citation_signature","submit_replication":"https://pith.science/pith/3D6EGCK4PR56BZBATZUU6P7PPE/action/replication_record"}},"created_at":"2026-05-18T00:06:28.623100+00:00","updated_at":"2026-05-18T00:06:28.623100+00:00"}