{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:S6PZHZKJILYAR4VCVVNCUJUAZ2","short_pith_number":"pith:S6PZHZKJ","schema_version":"1.0","canonical_sha256":"979f93e54942f008f2a2ad5a2a2680cea42f56a837bf530ac07c4d5b191e18be","source":{"kind":"arxiv","id":"2406.03791","version":1},"attestation_state":"computed","paper":{"title":"Speed of Light Exact Greedy Decoding for RNN-T Speech Recognition Models on GPU","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Daniel Galvez, Hainan Xu, Tim Kaldewey, Vladimir Bataev","submitted_at":"2024-06-06T07:03:28Z","abstract_excerpt":"The vast majority of inference time for RNN Transducer (RNN-T) models today is spent on decoding. Current state-of-the-art RNN-T decoding implementations leave the GPU idle ~80% of the time. Leveraging a new CUDA 12.4 feature, CUDA graph conditional nodes, we present an exact GPU-based implementation of greedy decoding for RNN-T models that eliminates this idle time. Our optimizations speed up a 1.1 billion parameter RNN-T model end-to-end by a factor of 2.5x. This technique can applied to the \"label looping\" alternative greedy decoding algorithm as well, achieving 1.7x and 1.4x end-to-end spe"},"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":"2406.03791","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T07:03:28Z","cross_cats_sorted":[],"title_canon_sha256":"4d16577e2b99fa90afc48e22ae2e570911bc2621274cb41bb1ff108b2be2437c","abstract_canon_sha256":"4519d445edf69f43aa0e195cbb5335aea21d60b9adb7de13d813e851ff87dca8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:28:15.961971Z","signature_b64":"PL0/mEKC3J0LS5f9AwDkhcL9DtL/ilL24o8E2TmEEJ7K7Kzo+/p5RJKCgJmUnN/Zoro9TteLvOLrdszh6zJODg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"979f93e54942f008f2a2ad5a2a2680cea42f56a837bf530ac07c4d5b191e18be","last_reissued_at":"2026-07-05T08:28:15.961593Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:28:15.961593Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Speed of Light Exact Greedy Decoding for RNN-T Speech Recognition Models on GPU","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Daniel Galvez, Hainan Xu, Tim Kaldewey, Vladimir Bataev","submitted_at":"2024-06-06T07:03:28Z","abstract_excerpt":"The vast majority of inference time for RNN Transducer (RNN-T) models today is spent on decoding. Current state-of-the-art RNN-T decoding implementations leave the GPU idle ~80% of the time. Leveraging a new CUDA 12.4 feature, CUDA graph conditional nodes, we present an exact GPU-based implementation of greedy decoding for RNN-T models that eliminates this idle time. Our optimizations speed up a 1.1 billion parameter RNN-T model end-to-end by a factor of 2.5x. This technique can applied to the \"label looping\" alternative greedy decoding algorithm as well, achieving 1.7x and 1.4x end-to-end spe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.03791","kind":"arxiv","version":1},"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/2406.03791/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":"2406.03791","created_at":"2026-07-05T08:28:15.961650+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.03791v1","created_at":"2026-07-05T08:28:15.961650+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.03791","created_at":"2026-07-05T08:28:15.961650+00:00"},{"alias_kind":"pith_short_12","alias_value":"S6PZHZKJILYA","created_at":"2026-07-05T08:28:15.961650+00:00"},{"alias_kind":"pith_short_16","alias_value":"S6PZHZKJILYAR4VC","created_at":"2026-07-05T08:28:15.961650+00:00"},{"alias_kind":"pith_short_8","alias_value":"S6PZHZKJ","created_at":"2026-07-05T08:28:15.961650+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2","json":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2.json","graph_json":"https://pith.science/api/pith-number/S6PZHZKJILYAR4VCVVNCUJUAZ2/graph.json","events_json":"https://pith.science/api/pith-number/S6PZHZKJILYAR4VCVVNCUJUAZ2/events.json","paper":"https://pith.science/paper/S6PZHZKJ"},"agent_actions":{"view_html":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2","download_json":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2.json","view_paper":"https://pith.science/paper/S6PZHZKJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.03791&json=true","fetch_graph":"https://pith.science/api/pith-number/S6PZHZKJILYAR4VCVVNCUJUAZ2/graph.json","fetch_events":"https://pith.science/api/pith-number/S6PZHZKJILYAR4VCVVNCUJUAZ2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2/action/storage_attestation","attest_author":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2/action/author_attestation","sign_citation":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2/action/citation_signature","submit_replication":"https://pith.science/pith/S6PZHZKJILYAR4VCVVNCUJUAZ2/action/replication_record"}},"created_at":"2026-07-05T08:28:15.961650+00:00","updated_at":"2026-07-05T08:28:15.961650+00:00"}