{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:545VV5TWOWRWQ3PEMEMSEYRLMS","short_pith_number":"pith:545VV5TW","schema_version":"1.0","canonical_sha256":"ef3b5af67675a3686de4611922622b6492d374cc5cd6c5e1665e4f3fd206ba91","source":{"kind":"arxiv","id":"1806.03377","version":1},"attestation_state":"computed","paper":{"title":"PipeDream: Fast and Efficient Pipeline Parallel DNN Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Aaron Harlap, Amar Phanishayee, Deepak Narayanan, Greg Ganger, Nikhil Devanur, Phil Gibbons, Vivek Seshadri","submitted_at":"2018-06-08T23:18:08Z","abstract_excerpt":"PipeDream is a Deep Neural Network(DNN) training system for GPUs that parallelizes computation by pipelining execution across multiple machines. Its pipeline parallel computing model avoids the slowdowns faced by data-parallel training when large models and/or limited network bandwidth induce high communication-to-computation ratios. PipeDream reduces communication by up to 95% for large DNNs relative to data-parallel training, and allows perfect overlap of communication and computation. PipeDream keeps all available GPUs productive by systematically partitioning DNN layers among them to balan"},"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":"1806.03377","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2018-06-08T23:18:08Z","cross_cats_sorted":[],"title_canon_sha256":"830960112e779d7ddea5ad989371da375eecfae6c6bc8a1d8c4c79fc80fcfe9d","abstract_canon_sha256":"157b828c47c63f7f239a1f169dfae3ac594b57e6578052e316685ea224ca23a4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:13:43.574928Z","signature_b64":"bqt1zIVlfxqFa4sHUH13X6yvSi/3ugKDGRMXsxttyKpda/hGZ+io+7Yi1+gQ1KAw4w7GoahtqllO+3WNxPQjAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ef3b5af67675a3686de4611922622b6492d374cc5cd6c5e1665e4f3fd206ba91","last_reissued_at":"2026-05-18T00:13:43.574235Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:13:43.574235Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PipeDream: Fast and Efficient Pipeline Parallel DNN Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Aaron Harlap, Amar Phanishayee, Deepak Narayanan, Greg Ganger, Nikhil Devanur, Phil Gibbons, Vivek Seshadri","submitted_at":"2018-06-08T23:18:08Z","abstract_excerpt":"PipeDream is a Deep Neural Network(DNN) training system for GPUs that parallelizes computation by pipelining execution across multiple machines. Its pipeline parallel computing model avoids the slowdowns faced by data-parallel training when large models and/or limited network bandwidth induce high communication-to-computation ratios. PipeDream reduces communication by up to 95% for large DNNs relative to data-parallel training, and allows perfect overlap of communication and computation. PipeDream keeps all available GPUs productive by systematically partitioning DNN layers among them to balan"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1806.03377","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":""},"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":"1806.03377","created_at":"2026-05-18T00:13:43.574335+00:00"},{"alias_kind":"arxiv_version","alias_value":"1806.03377v1","created_at":"2026-05-18T00:13:43.574335+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1806.03377","created_at":"2026-05-18T00:13:43.574335+00:00"},{"alias_kind":"pith_short_12","alias_value":"545VV5TWOWRW","created_at":"2026-05-18T12:32:05.422762+00:00"},{"alias_kind":"pith_short_16","alias_value":"545VV5TWOWRWQ3PE","created_at":"2026-05-18T12:32:05.422762+00:00"},{"alias_kind":"pith_short_8","alias_value":"545VV5TW","created_at":"2026-05-18T12:32:05.422762+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":6,"sample":[{"citing_arxiv_id":"2204.06745","citing_title":"GPT-NeoX-20B: An Open-Source Autoregressive Language Model","ref_index":34,"is_internal_anchor":true},{"citing_arxiv_id":"2408.12935","citing_title":"AI Safety Landscape for Large Language Models: Taxonomy, State-of-the-art, and Future Directions","ref_index":298,"is_internal_anchor":true},{"citing_arxiv_id":"2509.05276","citing_title":"SpikingBrain: Spiking Brain-inspired Large Models","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"1910.02054","citing_title":"ZeRO: Memory Optimizations Toward Training Trillion Parameter Models","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2512.16776","citing_title":"Kling-Omni Technical Report","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13434","citing_title":"Rescaled Asynchronous SGD: Optimal Distributed Optimization under Data and System Heterogeneity","ref_index":116,"is_internal_anchor":true},{"citing_arxiv_id":"2208.07339","citing_title":"LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale","ref_index":73,"is_internal_anchor":false},{"citing_arxiv_id":"2309.14509","citing_title":"DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2101.03961","citing_title":"Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2202.08906","citing_title":"ST-MoE: Designing Stable and Transferable Sparse Expert Models","ref_index":62,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27085","citing_title":"Efficient Training on Multiple Consumer GPUs with RoundPipe","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2304.11277","citing_title":"PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2401.06066","citing_title":"DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2006.16668","citing_title":"GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"1909.08053","citing_title":"Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2507.20534","citing_title":"Kimi K2: Open Agentic Intelligence","ref_index":22,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS","json":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS.json","graph_json":"https://pith.science/api/pith-number/545VV5TWOWRWQ3PEMEMSEYRLMS/graph.json","events_json":"https://pith.science/api/pith-number/545VV5TWOWRWQ3PEMEMSEYRLMS/events.json","paper":"https://pith.science/paper/545VV5TW"},"agent_actions":{"view_html":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS","download_json":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS.json","view_paper":"https://pith.science/paper/545VV5TW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1806.03377&json=true","fetch_graph":"https://pith.science/api/pith-number/545VV5TWOWRWQ3PEMEMSEYRLMS/graph.json","fetch_events":"https://pith.science/api/pith-number/545VV5TWOWRWQ3PEMEMSEYRLMS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS/action/storage_attestation","attest_author":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS/action/author_attestation","sign_citation":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS/action/citation_signature","submit_replication":"https://pith.science/pith/545VV5TWOWRWQ3PEMEMSEYRLMS/action/replication_record"}},"created_at":"2026-05-18T00:13:43.574335+00:00","updated_at":"2026-05-18T00:13:43.574335+00:00"}