{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:YMDGPZLLFCQNGQ7L74RDAWMF3T","short_pith_number":"pith:YMDGPZLL","schema_version":"1.0","canonical_sha256":"c30667e56b28a0d343ebff22305985dcd5b09a1775a65219c1fadbe278819e51","source":{"kind":"arxiv","id":"2309.06497","version":1},"attestation_state":"computed","paper":{"title":"A Distributed Data-Parallel PyTorch Implementation of the Distributed Shampoo Optimizer for Training Neural Networks At-Scale","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","cs.MS","math.OC"],"primary_cat":"cs.LG","authors_text":"Dheevatsa Mudigere, Hao-Jun Michael Shi, Jose Gallego-Posada, Kaushik Rangadurai, Michael Rabbat, Shintaro Iwasaki, Tsung-Hsien Lee, Zhijing Li","submitted_at":"2023-09-12T18:11:10Z","abstract_excerpt":"Shampoo is an online and stochastic optimization algorithm belonging to the AdaGrad family of methods for training neural networks. It constructs a block-diagonal preconditioner where each block consists of a coarse Kronecker product approximation to full-matrix AdaGrad for each parameter of the neural network. In this work, we provide a complete description of the algorithm as well as the performance optimizations that our implementation leverages to train deep networks at-scale in PyTorch. Our implementation enables fast multi-GPU distributed data-parallel training by distributing the memory"},"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":"2309.06497","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-09-12T18:11:10Z","cross_cats_sorted":["cs.DC","cs.MS","math.OC"],"title_canon_sha256":"618edec4cb070839a2daf4bbb58585343bdabb096e6ffee933dad5be1491c7af","abstract_canon_sha256":"59a96b33a0aed909bc606f2d8151b09295c29b45f0323a5319116d29d9606404"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:50:18.476696Z","signature_b64":"7B3voIKvJnWp6cyyeFRGWx9LDigl9d81ZZ4A59LUeuXH0VRqBfOSThz7bGHdn1djymWpvmvYmwIxqb34VBg7AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c30667e56b28a0d343ebff22305985dcd5b09a1775a65219c1fadbe278819e51","last_reissued_at":"2026-07-05T06:50:18.476190Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:50:18.476190Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Distributed Data-Parallel PyTorch Implementation of the Distributed Shampoo Optimizer for Training Neural Networks At-Scale","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","cs.MS","math.OC"],"primary_cat":"cs.LG","authors_text":"Dheevatsa Mudigere, Hao-Jun Michael Shi, Jose Gallego-Posada, Kaushik Rangadurai, Michael Rabbat, Shintaro Iwasaki, Tsung-Hsien Lee, Zhijing Li","submitted_at":"2023-09-12T18:11:10Z","abstract_excerpt":"Shampoo is an online and stochastic optimization algorithm belonging to the AdaGrad family of methods for training neural networks. It constructs a block-diagonal preconditioner where each block consists of a coarse Kronecker product approximation to full-matrix AdaGrad for each parameter of the neural network. In this work, we provide a complete description of the algorithm as well as the performance optimizations that our implementation leverages to train deep networks at-scale in PyTorch. Our implementation enables fast multi-GPU distributed data-parallel training by distributing the memory"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.06497","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/2309.06497/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":"2309.06497","created_at":"2026-07-05T06:50:18.476249+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.06497v1","created_at":"2026-07-05T06:50:18.476249+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.06497","created_at":"2026-07-05T06:50:18.476249+00:00"},{"alias_kind":"pith_short_12","alias_value":"YMDGPZLLFCQN","created_at":"2026-07-05T06:50:18.476249+00:00"},{"alias_kind":"pith_short_16","alias_value":"YMDGPZLLFCQNGQ7L","created_at":"2026-07-05T06:50:18.476249+00:00"},{"alias_kind":"pith_short_8","alias_value":"YMDGPZLL","created_at":"2026-07-05T06:50:18.476249+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05895","citing_title":"MatrixFSDP: communication-free matrix optimizers under ZeRO-3 parameter sharding","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2606.27153","citing_title":"DMuon: Efficient Distributed Muon Training with Near-Adam Overhead","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23357","citing_title":"SOAP-Bubbles: Structured Weight Uncertainty for Neural Networks","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06418","citing_title":"Double Preconditioning (DoPr): Optimization for Test-Time Performance, not Validation Loss","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04058","citing_title":"Spectral Scaling Laws of Muon","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02365","citing_title":"FOAM: Frequency and Operator Error-Based Adaptive Damping Method for Reducing Staleness-Oriented Error for Shampoo","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18106","citing_title":"Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers","ref_index":141,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16165","citing_title":"Second-Order Multi-Level Variance Correction for Modality Competition in Multimodal Models","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18106","citing_title":"Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers","ref_index":139,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16184","citing_title":"Runtime-Orchestrated Second-Order Optimization for Scalable LLM Training","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2409.20325","citing_title":"Old Optimizer, New Norm: An Anthology","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2603.20527","citing_title":"RMNP: Row-Momentum Normalized Preconditioning for Scalable Matrix-Based Optimization","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09552","citing_title":"Phases of Muon: When Muon Eclipses SignSGD","ref_index":66,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06316","citing_title":"Pro-KLShampoo: Projected KL-Shampoo with Whitening Recovered by Orthogonalization","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04418","citing_title":"Demystifying Manifold Constraints in LLM Pre-training","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T","json":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T.json","graph_json":"https://pith.science/api/pith-number/YMDGPZLLFCQNGQ7L74RDAWMF3T/graph.json","events_json":"https://pith.science/api/pith-number/YMDGPZLLFCQNGQ7L74RDAWMF3T/events.json","paper":"https://pith.science/paper/YMDGPZLL"},"agent_actions":{"view_html":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T","download_json":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T.json","view_paper":"https://pith.science/paper/YMDGPZLL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.06497&json=true","fetch_graph":"https://pith.science/api/pith-number/YMDGPZLLFCQNGQ7L74RDAWMF3T/graph.json","fetch_events":"https://pith.science/api/pith-number/YMDGPZLLFCQNGQ7L74RDAWMF3T/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T/action/storage_attestation","attest_author":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T/action/author_attestation","sign_citation":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T/action/citation_signature","submit_replication":"https://pith.science/pith/YMDGPZLLFCQNGQ7L74RDAWMF3T/action/replication_record"}},"created_at":"2026-07-05T06:50:18.476249+00:00","updated_at":"2026-07-05T06:50:18.476249+00:00"}