{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:K6DYAONSAFKLCGOLXMREAQRJ4M","short_pith_number":"pith:K6DYAONS","schema_version":"1.0","canonical_sha256":"57878039b20154b119cbbb22404229e32c9c5d7b7b9b5a3a93f925af43164e1c","source":{"kind":"arxiv","id":"2406.08552","version":2},"attestation_state":"computed","paper":{"title":"DiTFastAttn: Attention Compression for Diffusion Transformer Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Guohao Dai, Hanling Zhang, Linfeng Zhang, Pu Lu, Shengen Yan, Tianchen Zhao, Xuefei Ning, Yu Wang, Zhihang Yuan","submitted_at":"2024-06-12T18:00:08Z","abstract_excerpt":"Diffusion Transformers (DiT) excel at image and video generation but face computational challenges due to the quadratic complexity of self-attention operators. We propose DiTFastAttn, a post-training compression method to alleviate the computational bottleneck of DiT. We identify three key redundancies in the attention computation during DiT inference: (1) spatial redundancy, where many attention heads focus on local information; (2) temporal redundancy, with high similarity between the attention outputs of neighboring steps; (3) conditional redundancy, where conditional and unconditional infe"},"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":"2406.08552","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-06-12T18:00:08Z","cross_cats_sorted":[],"title_canon_sha256":"0f7811ea2fcac554bf97403fe27a28683cde400322bd4152a6034b5213a3b24d","abstract_canon_sha256":"42008817e8d006560fddfdb0225bbc7edba74d2e1c725abebf0b3901ff8f6526"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:22:19.289372Z","signature_b64":"rklioxDSe8zbKaK8NUj7HJEjfC0yYTvlfiJmxHyFB3z8eqhBDy00U7NjsQ9JOHAh+4rpXTh5jqFOtsi0m1h3DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"57878039b20154b119cbbb22404229e32c9c5d7b7b9b5a3a93f925af43164e1c","last_reissued_at":"2026-07-05T09:22:19.288881Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:22:19.288881Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DiTFastAttn: Attention Compression for Diffusion Transformer Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Guohao Dai, Hanling Zhang, Linfeng Zhang, Pu Lu, Shengen Yan, Tianchen Zhao, Xuefei Ning, Yu Wang, Zhihang Yuan","submitted_at":"2024-06-12T18:00:08Z","abstract_excerpt":"Diffusion Transformers (DiT) excel at image and video generation but face computational challenges due to the quadratic complexity of self-attention operators. We propose DiTFastAttn, a post-training compression method to alleviate the computational bottleneck of DiT. We identify three key redundancies in the attention computation during DiT inference: (1) spatial redundancy, where many attention heads focus on local information; (2) temporal redundancy, with high similarity between the attention outputs of neighboring steps; (3) conditional redundancy, where conditional and unconditional infe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.08552","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/2406.08552/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":"2406.08552","created_at":"2026-07-05T09:22:19.288942+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.08552v2","created_at":"2026-07-05T09:22:19.288942+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.08552","created_at":"2026-07-05T09:22:19.288942+00:00"},{"alias_kind":"pith_short_12","alias_value":"K6DYAONSAFKL","created_at":"2026-07-05T09:22:19.288942+00:00"},{"alias_kind":"pith_short_16","alias_value":"K6DYAONSAFKLCGOL","created_at":"2026-07-05T09:22:19.288942+00:00"},{"alias_kind":"pith_short_8","alias_value":"K6DYAONS","created_at":"2026-07-05T09:22:19.288942+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26632","citing_title":"RT-Lynx: Putting the GEMM Sparsity In a Right Way for Diffusion Models","ref_index":71,"is_internal_anchor":false},{"citing_arxiv_id":"2405.14430","citing_title":"PipeFusion: Patch-level Pipeline Parallelism for Diffusion Transformers Inference","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2603.28049","citing_title":"Drift-AR: Single-Step Visual Autoregressive Generation via Anti-Symmetric Drifting","ref_index":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M","json":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M.json","graph_json":"https://pith.science/api/pith-number/K6DYAONSAFKLCGOLXMREAQRJ4M/graph.json","events_json":"https://pith.science/api/pith-number/K6DYAONSAFKLCGOLXMREAQRJ4M/events.json","paper":"https://pith.science/paper/K6DYAONS"},"agent_actions":{"view_html":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M","download_json":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M.json","view_paper":"https://pith.science/paper/K6DYAONS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.08552&json=true","fetch_graph":"https://pith.science/api/pith-number/K6DYAONSAFKLCGOLXMREAQRJ4M/graph.json","fetch_events":"https://pith.science/api/pith-number/K6DYAONSAFKLCGOLXMREAQRJ4M/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M/action/timestamp_anchor","attest_storage":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M/action/storage_attestation","attest_author":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M/action/author_attestation","sign_citation":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M/action/citation_signature","submit_replication":"https://pith.science/pith/K6DYAONSAFKLCGOLXMREAQRJ4M/action/replication_record"}},"created_at":"2026-07-05T09:22:19.288942+00:00","updated_at":"2026-07-05T09:22:19.288942+00:00"}