{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:EMFK5LF3JUIUBNTYQT4YPOKHFD","short_pith_number":"pith:EMFK5LF3","canonical_record":{"source":{"id":"2412.03317","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-04T13:52:04Z","cross_cats_sorted":[],"title_canon_sha256":"d5bd41fa6c5a169292881c8f64726cab7e43e6b5780e5373e6eb0a4a084bbfe6","abstract_canon_sha256":"6b72afec1eef72029103fc8c01d04cc0a3dd060a2891973db15e5b9963ed3f82"},"schema_version":"1.0"},"canonical_sha256":"230aaeacbb4d1140b67884f987b94728f8886b6652618b886fd564fee5ac47eb","source":{"kind":"arxiv","id":"2412.03317","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.03317","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"arxiv_version","alias_value":"2412.03317v2","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.03317","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"pith_short_12","alias_value":"EMFK5LF3JUIU","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"pith_short_16","alias_value":"EMFK5LF3JUIUBNTY","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"pith_short_8","alias_value":"EMFK5LF3","created_at":"2026-07-05T10:02:55Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:EMFK5LF3JUIUBNTYQT4YPOKHFD","target":"record","payload":{"canonical_record":{"source":{"id":"2412.03317","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-04T13:52:04Z","cross_cats_sorted":[],"title_canon_sha256":"d5bd41fa6c5a169292881c8f64726cab7e43e6b5780e5373e6eb0a4a084bbfe6","abstract_canon_sha256":"6b72afec1eef72029103fc8c01d04cc0a3dd060a2891973db15e5b9963ed3f82"},"schema_version":"1.0"},"canonical_sha256":"230aaeacbb4d1140b67884f987b94728f8886b6652618b886fd564fee5ac47eb","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:02:55.628905Z","signature_b64":"XQ1Fnl/Rz64qcOf9VvZdNCweYzoGsxLea344F6CLKnO3wsLPxBIHbDpQ1ENW1U3KWvr09S5ggpAYKZr3YRLFDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"230aaeacbb4d1140b67884f987b94728f8886b6652618b886fd564fee5ac47eb","last_reissued_at":"2026-07-05T10:02:55.628472Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:02:55.628472Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2412.03317","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:02:55Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6k5akWA3XbQ/QPcd8ok4MECT8I/eDsIWhHVdsA6AKVMJDHve5UI9F2LpoSyd7x2fuiFO7wTlWbziireZh6meDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T08:49:33.530150Z"},"content_sha256":"8b403c971c1bf9f0fb61491c59f1e6e6dc5a6a68a59a00e68c92288f77b48adf","schema_version":"1.0","event_id":"sha256:8b403c971c1bf9f0fb61491c59f1e6e6dc5a6a68a59a00e68c92288f77b48adf"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:EMFK5LF3JUIUBNTYQT4YPOKHFD","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"FlashAttention on a Napkin: A Diagrammatic Approach to Deep Learning IO-Awareness","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Gioele Zardini, Vincent Abbott","submitted_at":"2024-12-04T13:52:04Z","abstract_excerpt":"Optimizing deep learning algorithms currently requires slow, manual derivation, potentially leaving much performance untapped. Methods like FlashAttention have achieved a x6 performance improvement over native PyTorch by avoiding unnecessary data transfers, but required three iterations over three years to be developed. Automated compiled methods have consistently lagged behind. This paper extends Neural Circuit Diagrams for deep learning models to consider resource usage and the distribution of tasks across a GPU hierarchy. We show how diagrams can use simple relabellings to derive high-level"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.03317","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/2412.03317/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:02:55Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8tRTnFK3KNnyLu3rsXGWO+G6JZOxZXCEFGGKgs8GIS9I4HsghNM5ozNO9r2ANiCzpEb6KOiZToQ5qhtNZyX2CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T08:49:33.531542Z"},"content_sha256":"7cf3b0e145d06c6d4e4a16f3363b0c72dd391d78bb7ab975b321cf2dd3cc981f","schema_version":"1.0","event_id":"sha256:7cf3b0e145d06c6d4e4a16f3363b0c72dd391d78bb7ab975b321cf2dd3cc981f"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/EMFK5LF3JUIUBNTYQT4YPOKHFD/bundle.json","state_url":"https://pith.science/pith/EMFK5LF3JUIUBNTYQT4YPOKHFD/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/EMFK5LF3JUIUBNTYQT4YPOKHFD/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-08T08:49:33Z","links":{"resolver":"https://pith.science/pith/EMFK5LF3JUIUBNTYQT4YPOKHFD","bundle":"https://pith.science/pith/EMFK5LF3JUIUBNTYQT4YPOKHFD/bundle.json","state":"https://pith.science/pith/EMFK5LF3JUIUBNTYQT4YPOKHFD/state.json","well_known_bundle":"https://pith.science/.well-known/pith/EMFK5LF3JUIUBNTYQT4YPOKHFD/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:EMFK5LF3JUIUBNTYQT4YPOKHFD","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"6b72afec1eef72029103fc8c01d04cc0a3dd060a2891973db15e5b9963ed3f82","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-04T13:52:04Z","title_canon_sha256":"d5bd41fa6c5a169292881c8f64726cab7e43e6b5780e5373e6eb0a4a084bbfe6"},"schema_version":"1.0","source":{"id":"2412.03317","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.03317","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"arxiv_version","alias_value":"2412.03317v2","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.03317","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"pith_short_12","alias_value":"EMFK5LF3JUIU","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"pith_short_16","alias_value":"EMFK5LF3JUIUBNTY","created_at":"2026-07-05T10:02:55Z"},{"alias_kind":"pith_short_8","alias_value":"EMFK5LF3","created_at":"2026-07-05T10:02:55Z"}],"graph_snapshots":[{"event_id":"sha256:7cf3b0e145d06c6d4e4a16f3363b0c72dd391d78bb7ab975b321cf2dd3cc981f","target":"graph","created_at":"2026-07-05T10:02:55Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2412.03317/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Optimizing deep learning algorithms currently requires slow, manual derivation, potentially leaving much performance untapped. Methods like FlashAttention have achieved a x6 performance improvement over native PyTorch by avoiding unnecessary data transfers, but required three iterations over three years to be developed. Automated compiled methods have consistently lagged behind. This paper extends Neural Circuit Diagrams for deep learning models to consider resource usage and the distribution of tasks across a GPU hierarchy. We show how diagrams can use simple relabellings to derive high-level","authors_text":"Gioele Zardini, Vincent Abbott","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-04T13:52:04Z","title":"FlashAttention on a Napkin: A Diagrammatic Approach to Deep Learning IO-Awareness"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.03317","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:8b403c971c1bf9f0fb61491c59f1e6e6dc5a6a68a59a00e68c92288f77b48adf","target":"record","created_at":"2026-07-05T10:02:55Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"6b72afec1eef72029103fc8c01d04cc0a3dd060a2891973db15e5b9963ed3f82","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-04T13:52:04Z","title_canon_sha256":"d5bd41fa6c5a169292881c8f64726cab7e43e6b5780e5373e6eb0a4a084bbfe6"},"schema_version":"1.0","source":{"id":"2412.03317","kind":"arxiv","version":2}},"canonical_sha256":"230aaeacbb4d1140b67884f987b94728f8886b6652618b886fd564fee5ac47eb","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"230aaeacbb4d1140b67884f987b94728f8886b6652618b886fd564fee5ac47eb","first_computed_at":"2026-07-05T10:02:55.628472Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:02:55.628472Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"XQ1Fnl/Rz64qcOf9VvZdNCweYzoGsxLea344F6CLKnO3wsLPxBIHbDpQ1ENW1U3KWvr09S5ggpAYKZr3YRLFDA==","signature_status":"signed_v1","signed_at":"2026-07-05T10:02:55.628905Z","signed_message":"canonical_sha256_bytes"},"source_id":"2412.03317","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:8b403c971c1bf9f0fb61491c59f1e6e6dc5a6a68a59a00e68c92288f77b48adf","sha256:7cf3b0e145d06c6d4e4a16f3363b0c72dd391d78bb7ab975b321cf2dd3cc981f"],"state_sha256":"267bed8fd8771d8343fe90a5fae1e3d7b5d3c09f32aa425afb9623c46161bbe2"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7KQZAGsdhwih7F8vLV37Se7TvKdjQE6S8xmSLM4Z0VRKef6bpMvi//IaJTEUNGEyian4Gw8AvzudoV4EYonaBA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-08T08:49:33.538979Z","bundle_sha256":"34d57b68e046b7c27a5d1ff9c512e882d497193ae73d5746f8dca62c941fb78f"}}