{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:AR7B3DEVMOTSBRCCVXUCLYSRRE","short_pith_number":"pith:AR7B3DEV","schema_version":"1.0","canonical_sha256":"047e1d8c9563a720c442ade825e25189007290160bedcc76b7e523bfd71812d3","source":{"kind":"arxiv","id":"2401.14489","version":2},"attestation_state":"computed","paper":{"title":"The Case for Co-Designing Model Architectures with Hardware","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DC","authors_text":"Aamir Shafi, Deepak Narayanan, Dhabaleswar Panda, Hari Subramoni, Jacob Hatef, Junqi Yin, Quentin Anthony, Stas Bekman, Stella Biderman","submitted_at":"2024-01-25T19:50:31Z","abstract_excerpt":"While GPUs are responsible for training the vast majority of state-of-the-art deep learning models, the implications of their architecture are often overlooked when designing new deep learning (DL) models. As a consequence, modifying a DL model to be more amenable to the target hardware can significantly improve the runtime performance of DL training and inference. In this paper, we provide a set of guidelines for users to maximize the runtime performance of their transformer models. These guidelines have been created by carefully considering the impact of various model hyperparameters control"},"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":"2401.14489","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2024-01-25T19:50:31Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"93d2e3b81af83d220120cd555e3c538adbf44cf463d598b7ba69e46ae4c866e0","abstract_canon_sha256":"2b0354a80ba906644ad0e9b3f6df5f5f0384ddf4563f84d4fe2ad3ce8a7afd54"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:39:34.446811Z","signature_b64":"vFpe881wbzIvvQAawnvz2cb9vbbL6c83HVzq8YXbqwVCJMN6PjIK241fJ8/g43MXcvlp8rRWW2t/0pjT9dvxDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"047e1d8c9563a720c442ade825e25189007290160bedcc76b7e523bfd71812d3","last_reissued_at":"2026-07-05T07:39:34.446332Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:39:34.446332Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Case for Co-Designing Model Architectures with Hardware","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DC","authors_text":"Aamir Shafi, Deepak Narayanan, Dhabaleswar Panda, Hari Subramoni, Jacob Hatef, Junqi Yin, Quentin Anthony, Stas Bekman, Stella Biderman","submitted_at":"2024-01-25T19:50:31Z","abstract_excerpt":"While GPUs are responsible for training the vast majority of state-of-the-art deep learning models, the implications of their architecture are often overlooked when designing new deep learning (DL) models. As a consequence, modifying a DL model to be more amenable to the target hardware can significantly improve the runtime performance of DL training and inference. In this paper, we provide a set of guidelines for users to maximize the runtime performance of their transformer models. These guidelines have been created by carefully considering the impact of various model hyperparameters control"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.14489","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/2401.14489/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":"2401.14489","created_at":"2026-07-05T07:39:34.446402+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.14489v2","created_at":"2026-07-05T07:39:34.446402+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.14489","created_at":"2026-07-05T07:39:34.446402+00:00"},{"alias_kind":"pith_short_12","alias_value":"AR7B3DEVMOTS","created_at":"2026-07-05T07:39:34.446402+00:00"},{"alias_kind":"pith_short_16","alias_value":"AR7B3DEVMOTSBRCC","created_at":"2026-07-05T07:39:34.446402+00:00"},{"alias_kind":"pith_short_8","alias_value":"AR7B3DEV","created_at":"2026-07-05T07:39:34.446402+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2412.13663","citing_title":"Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference","ref_index":115,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE","json":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE.json","graph_json":"https://pith.science/api/pith-number/AR7B3DEVMOTSBRCCVXUCLYSRRE/graph.json","events_json":"https://pith.science/api/pith-number/AR7B3DEVMOTSBRCCVXUCLYSRRE/events.json","paper":"https://pith.science/paper/AR7B3DEV"},"agent_actions":{"view_html":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE","download_json":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE.json","view_paper":"https://pith.science/paper/AR7B3DEV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.14489&json=true","fetch_graph":"https://pith.science/api/pith-number/AR7B3DEVMOTSBRCCVXUCLYSRRE/graph.json","fetch_events":"https://pith.science/api/pith-number/AR7B3DEVMOTSBRCCVXUCLYSRRE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE/action/storage_attestation","attest_author":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE/action/author_attestation","sign_citation":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE/action/citation_signature","submit_replication":"https://pith.science/pith/AR7B3DEVMOTSBRCCVXUCLYSRRE/action/replication_record"}},"created_at":"2026-07-05T07:39:34.446402+00:00","updated_at":"2026-07-05T07:39:34.446402+00:00"}