{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:2VH3MOWI7YRBP3TL7S7MLJQVFI","short_pith_number":"pith:2VH3MOWI","schema_version":"1.0","canonical_sha256":"d54fb63ac8fe2217ee6bfcbec5a6152a19e24ebf966785c6da9eda56cb6f277e","source":{"kind":"arxiv","id":"2405.16256","version":2},"attestation_state":"computed","paper":{"title":"HETHUB: A Distributed Training System with Heterogeneous Cluster for Large-Scale Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DC","authors_text":"Chunsheng Shui, Guohao Dai, Haolin Ye, Hao Zhang, Quanlu Zhang, Shengen Yan, Sheng Wang, Sipei Gu, Si Xu, Xuefei Ning, Yan Zeng, Yu Wang, Zhezheng Lin, Zixiao Huang","submitted_at":"2024-05-25T14:36:35Z","abstract_excerpt":"Training large-scale models relies on a vast number of computing resources. For example, training the GPT-4 model (1.8 trillion parameters) requires 25000 A100 GPUs . It is a challenge to build a large-scale cluster with one type of GPU-accelerator. Using multiple types of GPU-accelerators to construct a large-scale cluster is an effective way to solve the problem of insufficient homogeneous GPU-accelerators. However, the existing distributed training systems for large-scale models only support homogeneous GPU-accelerators, not support heterogeneous GPU-accelerators. To address the problem, 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":"2405.16256","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2024-05-25T14:36:35Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0191edb7833f6726cacf149eddbac5f659de11815e1a87a6c4d33150dc367caa","abstract_canon_sha256":"3b8a38b61380671db42c95c8d56299c72d54b082feeba735a64f4baa346890a8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:53:40.487551Z","signature_b64":"4yhvm1IYiUZOfl7uTuvNC8zOMD9ncVRtJdrG3pdbI5xbGKe62fgvOjtgOEb5f1bXx/dxN/RCkpjwD4SrtSIjBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d54fb63ac8fe2217ee6bfcbec5a6152a19e24ebf966785c6da9eda56cb6f277e","last_reissued_at":"2026-07-05T08:53:40.487039Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:53:40.487039Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HETHUB: A Distributed Training System with Heterogeneous Cluster for Large-Scale Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DC","authors_text":"Chunsheng Shui, Guohao Dai, Haolin Ye, Hao Zhang, Quanlu Zhang, Shengen Yan, Sheng Wang, Sipei Gu, Si Xu, Xuefei Ning, Yan Zeng, Yu Wang, Zhezheng Lin, Zixiao Huang","submitted_at":"2024-05-25T14:36:35Z","abstract_excerpt":"Training large-scale models relies on a vast number of computing resources. For example, training the GPT-4 model (1.8 trillion parameters) requires 25000 A100 GPUs . It is a challenge to build a large-scale cluster with one type of GPU-accelerator. Using multiple types of GPU-accelerators to construct a large-scale cluster is an effective way to solve the problem of insufficient homogeneous GPU-accelerators. However, the existing distributed training systems for large-scale models only support homogeneous GPU-accelerators, not support heterogeneous GPU-accelerators. To address the problem, th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.16256","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/2405.16256/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":"2405.16256","created_at":"2026-07-05T08:53:40.487099+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.16256v2","created_at":"2026-07-05T08:53:40.487099+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.16256","created_at":"2026-07-05T08:53:40.487099+00:00"},{"alias_kind":"pith_short_12","alias_value":"2VH3MOWI7YRB","created_at":"2026-07-05T08:53:40.487099+00:00"},{"alias_kind":"pith_short_16","alias_value":"2VH3MOWI7YRBP3TL","created_at":"2026-07-05T08:53:40.487099+00:00"},{"alias_kind":"pith_short_8","alias_value":"2VH3MOWI","created_at":"2026-07-05T08:53:40.487099+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.17548","citing_title":"H2:Towards Efficient Large-Scale LLM Training on Hyper-Heterogeneous Cluster over 1,000 Chips","ref_index":40,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI","json":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI.json","graph_json":"https://pith.science/api/pith-number/2VH3MOWI7YRBP3TL7S7MLJQVFI/graph.json","events_json":"https://pith.science/api/pith-number/2VH3MOWI7YRBP3TL7S7MLJQVFI/events.json","paper":"https://pith.science/paper/2VH3MOWI"},"agent_actions":{"view_html":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI","download_json":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI.json","view_paper":"https://pith.science/paper/2VH3MOWI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.16256&json=true","fetch_graph":"https://pith.science/api/pith-number/2VH3MOWI7YRBP3TL7S7MLJQVFI/graph.json","fetch_events":"https://pith.science/api/pith-number/2VH3MOWI7YRBP3TL7S7MLJQVFI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI/action/storage_attestation","attest_author":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI/action/author_attestation","sign_citation":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI/action/citation_signature","submit_replication":"https://pith.science/pith/2VH3MOWI7YRBP3TL7S7MLJQVFI/action/replication_record"}},"created_at":"2026-07-05T08:53:40.487099+00:00","updated_at":"2026-07-05T08:53:40.487099+00:00"}