{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LOYYHJXGUAHVIZJBIEZZNKB42K","short_pith_number":"pith:LOYYHJXG","schema_version":"1.0","canonical_sha256":"5bb183a6e6a00f546521413396a83cd28ed2093ed113161d8c096f9001e85363","source":{"kind":"arxiv","id":"2401.09149","version":3},"attestation_state":"computed","paper":{"title":"InternEvo: Efficient Long-sequence Large Language Model Training via Hybrid Parallelism and Redundant Sharding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Diandian Gu, Guoteng Wang, Peng Sun, Qiaoling Chen, Qinghao Hu, Tianwei Zhang, Ting Huang, Xin Jin, Xun Chen, Yingtong Xiong, Yonggang Wen","submitted_at":"2024-01-17T11:47:59Z","abstract_excerpt":"Large language models (LLMs) with long sequences begin to power more and more fundamentally new applications we use every day. Existing methods for long-sequence LLM training are neither efficient nor compatible with commonly-used training algorithms such as FlashAttention. We design InternEvo to address these issues. InternEvo decouples all of the sharding dimensions into a new hierarchical space, and systematically analyzes the memory and communication cost of LLM training. Then, it generates an effective hybrid parallelism strategy. We design a new selective overlap mechanism to mitigate th"},"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":"2401.09149","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2024-01-17T11:47:59Z","cross_cats_sorted":[],"title_canon_sha256":"797785b0998bdcf271fc638d3a3e7b606268f72f6d1619732bb3ec043a24e703","abstract_canon_sha256":"c2536bc15688faf1c290c650afef2f178baf7161bbe152ab7b1b7194db581e4b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:36:05.959799Z","signature_b64":"2xqR+qwKzOR9Y445fE8RSIGwfYzILEwj1igIAh36pAc8PRE6jdsjlj2kRp+s/OJTqFPhGsDtYphT5rL9ZuB+Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5bb183a6e6a00f546521413396a83cd28ed2093ed113161d8c096f9001e85363","last_reissued_at":"2026-07-05T07:36:05.959321Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:36:05.959321Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"InternEvo: Efficient Long-sequence Large Language Model Training via Hybrid Parallelism and Redundant Sharding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Diandian Gu, Guoteng Wang, Peng Sun, Qiaoling Chen, Qinghao Hu, Tianwei Zhang, Ting Huang, Xin Jin, Xun Chen, Yingtong Xiong, Yonggang Wen","submitted_at":"2024-01-17T11:47:59Z","abstract_excerpt":"Large language models (LLMs) with long sequences begin to power more and more fundamentally new applications we use every day. Existing methods for long-sequence LLM training are neither efficient nor compatible with commonly-used training algorithms such as FlashAttention. We design InternEvo to address these issues. InternEvo decouples all of the sharding dimensions into a new hierarchical space, and systematically analyzes the memory and communication cost of LLM training. Then, it generates an effective hybrid parallelism strategy. We design a new selective overlap mechanism to mitigate th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.09149","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/2401.09149/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":"2401.09149","created_at":"2026-07-05T07:36:05.959363+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.09149v3","created_at":"2026-07-05T07:36:05.959363+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.09149","created_at":"2026-07-05T07:36:05.959363+00:00"},{"alias_kind":"pith_short_12","alias_value":"LOYYHJXGUAHV","created_at":"2026-07-05T07:36:05.959363+00:00"},{"alias_kind":"pith_short_16","alias_value":"LOYYHJXGUAHVIZJB","created_at":"2026-07-05T07:36:05.959363+00:00"},{"alias_kind":"pith_short_8","alias_value":"LOYYHJXG","created_at":"2026-07-05T07:36:05.959363+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2403.17297","citing_title":"InternLM2 Technical Report","ref_index":119,"is_internal_anchor":false},{"citing_arxiv_id":"2403.17297","citing_title":"InternLM2 Technical Report","ref_index":144,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07569","citing_title":"HexiSeq: Accommodating Long Context Training of LLMs over Heterogeneous Hardware","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2504.10479","citing_title":"InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K","json":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K.json","graph_json":"https://pith.science/api/pith-number/LOYYHJXGUAHVIZJBIEZZNKB42K/graph.json","events_json":"https://pith.science/api/pith-number/LOYYHJXGUAHVIZJBIEZZNKB42K/events.json","paper":"https://pith.science/paper/LOYYHJXG"},"agent_actions":{"view_html":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K","download_json":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K.json","view_paper":"https://pith.science/paper/LOYYHJXG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.09149&json=true","fetch_graph":"https://pith.science/api/pith-number/LOYYHJXGUAHVIZJBIEZZNKB42K/graph.json","fetch_events":"https://pith.science/api/pith-number/LOYYHJXGUAHVIZJBIEZZNKB42K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K/action/storage_attestation","attest_author":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K/action/author_attestation","sign_citation":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K/action/citation_signature","submit_replication":"https://pith.science/pith/LOYYHJXGUAHVIZJBIEZZNKB42K/action/replication_record"}},"created_at":"2026-07-05T07:36:05.959363+00:00","updated_at":"2026-07-05T07:36:05.959363+00:00"}