{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:NOOOAGJC7737FRV6TZFZTS4TQM","short_pith_number":"pith:NOOOAGJC","schema_version":"1.0","canonical_sha256":"6b9ce01922fff7f2c6be9e4b99cb938306a386ed9cf621ceca94f7746d4d91c6","source":{"kind":"arxiv","id":"1911.04610","version":3},"attestation_state":"computed","paper":{"title":"XPipe: Efficient Pipeline Model Parallelism for Multi-GPU DNN Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.PF"],"primary_cat":"cs.LG","authors_text":"Dongsheng Li, Lei Guan, Wotao Yin, Xicheng Lu","submitted_at":"2019-10-24T00:13:54Z","abstract_excerpt":"We propose XPipe, an efficient asynchronous pipeline model parallelism approach for multi-GPU DNN training. XPipe is designed to use multiple GPUs to concurrently and continuously train different parts of a DNN model. To improve GPU utilization and achieve high throughput, it splits a mini-batch into a set of micro-batches. It allows the overlapping of the pipelines of multiple micro-batches, including those belonging to different mini-batches. Most importantly, the novel weight prediction strategy adopted by XPipe enables it to effectively address the weight inconsistency and staleness issues"},"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":"1911.04610","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-10-24T00:13:54Z","cross_cats_sorted":["cs.PF"],"title_canon_sha256":"9fbe131b37d0bb5395bc27905c9e43863110cf5fe2c29d6a405dbf493e89659f","abstract_canon_sha256":"2bc152c6e93c551bec3bd7922471176e4feedce7dd50d6dab5d33b9f287313d1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:49:48.339382Z","signature_b64":"21OIbmQ3LXkU94/c6KWNnrZE5xkcZNqSgVUSVpEWSX82dIhQ5fa7FINkcXoGyXsCNhJWByNW2GDtV6gDN5f/AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6b9ce01922fff7f2c6be9e4b99cb938306a386ed9cf621ceca94f7746d4d91c6","last_reissued_at":"2026-07-05T01:49:48.338901Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:49:48.338901Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"XPipe: Efficient Pipeline Model Parallelism for Multi-GPU DNN Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.PF"],"primary_cat":"cs.LG","authors_text":"Dongsheng Li, Lei Guan, Wotao Yin, Xicheng Lu","submitted_at":"2019-10-24T00:13:54Z","abstract_excerpt":"We propose XPipe, an efficient asynchronous pipeline model parallelism approach for multi-GPU DNN training. XPipe is designed to use multiple GPUs to concurrently and continuously train different parts of a DNN model. To improve GPU utilization and achieve high throughput, it splits a mini-batch into a set of micro-batches. It allows the overlapping of the pipelines of multiple micro-batches, including those belonging to different mini-batches. Most importantly, the novel weight prediction strategy adopted by XPipe enables it to effectively address the weight inconsistency and staleness issues"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1911.04610","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/1911.04610/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":"1911.04610","created_at":"2026-07-05T01:49:48.338952+00:00"},{"alias_kind":"arxiv_version","alias_value":"1911.04610v3","created_at":"2026-07-05T01:49:48.338952+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1911.04610","created_at":"2026-07-05T01:49:48.338952+00:00"},{"alias_kind":"pith_short_12","alias_value":"NOOOAGJC7737","created_at":"2026-07-05T01:49:48.338952+00:00"},{"alias_kind":"pith_short_16","alias_value":"NOOOAGJC7737FRV6","created_at":"2026-07-05T01:49:48.338952+00:00"},{"alias_kind":"pith_short_8","alias_value":"NOOOAGJC","created_at":"2026-07-05T01:49:48.338952+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07881","citing_title":"Breaking the Bubble: Asynchronous Pipeline Parallel Training with Bounded Weight Inconsistency","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30634","citing_title":"One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM Pretraining","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27085","citing_title":"Efficient Training on Multiple Consumer GPUs with RoundPipe","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM","json":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM.json","graph_json":"https://pith.science/api/pith-number/NOOOAGJC7737FRV6TZFZTS4TQM/graph.json","events_json":"https://pith.science/api/pith-number/NOOOAGJC7737FRV6TZFZTS4TQM/events.json","paper":"https://pith.science/paper/NOOOAGJC"},"agent_actions":{"view_html":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM","download_json":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM.json","view_paper":"https://pith.science/paper/NOOOAGJC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1911.04610&json=true","fetch_graph":"https://pith.science/api/pith-number/NOOOAGJC7737FRV6TZFZTS4TQM/graph.json","fetch_events":"https://pith.science/api/pith-number/NOOOAGJC7737FRV6TZFZTS4TQM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM/action/storage_attestation","attest_author":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM/action/author_attestation","sign_citation":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM/action/citation_signature","submit_replication":"https://pith.science/pith/NOOOAGJC7737FRV6TZFZTS4TQM/action/replication_record"}},"created_at":"2026-07-05T01:49:48.338952+00:00","updated_at":"2026-07-05T01:49:48.338952+00:00"}