{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:UHAXZSIEIUGAP5YCZ2WXTWMK2B","short_pith_number":"pith:UHAXZSIE","schema_version":"1.0","canonical_sha256":"a1c17cc904450c07f702cead79d98ad05aecc67543904ed5ba6c3307e8e994aa","source":{"kind":"arxiv","id":"2509.00217","version":1},"attestation_state":"computed","paper":{"title":"Learning to Shard: RL for Co-optimizing the Parallelism Degrees and Per-operator Sharding Dimensions in Distributed LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Apala Guha, Hokchhay Tann, Preyas Shah, Ruokai Yin, Sattwik Deb Mishra, Xuan Zuo","submitted_at":"2025-08-29T20:01:35Z","abstract_excerpt":"Distributed LLM inference requires careful coordination of parallelization strategies across hundreds to thousands of NPUs to meet production SLOs. Current systems like Megatron-LM rely on static heuristics that separately configure parallelism degrees and per-operator sharding dimensions, leaving significant performance on the table as models scale and hardware topologies diversify. We introduce Learn to Shard, to our knowledge, the first RL-based approach to co-optimize both coarse-grained parallelism degrees and fine-grained per-operator sharding dimensions for distributed LLM inference. Ou"},"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":"2509.00217","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-29T20:01:35Z","cross_cats_sorted":["cs.DC"],"title_canon_sha256":"ee645f80179b4848125bb43063a7c6595c0cc189f13dd8e3aec7f07d9a6ab51c","abstract_canon_sha256":"fe4fb341aef5c9176af1b942c7aabb91706716fa93f088bef3a304d6b90e7ec3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:02:11.689187Z","signature_b64":"TG+ioT7I9gA4/EbP2Ry94vEaDOqk0duGwUrvyp84e1IxntforqAStd86B/p1evhjdO/ClG4QmydLEdsv25mABA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a1c17cc904450c07f702cead79d98ad05aecc67543904ed5ba6c3307e8e994aa","last_reissued_at":"2026-07-05T12:02:11.688712Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:02:11.688712Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning to Shard: RL for Co-optimizing the Parallelism Degrees and Per-operator Sharding Dimensions in Distributed LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Apala Guha, Hokchhay Tann, Preyas Shah, Ruokai Yin, Sattwik Deb Mishra, Xuan Zuo","submitted_at":"2025-08-29T20:01:35Z","abstract_excerpt":"Distributed LLM inference requires careful coordination of parallelization strategies across hundreds to thousands of NPUs to meet production SLOs. Current systems like Megatron-LM rely on static heuristics that separately configure parallelism degrees and per-operator sharding dimensions, leaving significant performance on the table as models scale and hardware topologies diversify. We introduce Learn to Shard, to our knowledge, the first RL-based approach to co-optimize both coarse-grained parallelism degrees and fine-grained per-operator sharding dimensions for distributed LLM inference. Ou"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.00217","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/2509.00217/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":"2509.00217","created_at":"2026-07-05T12:02:11.688773+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.00217v1","created_at":"2026-07-05T12:02:11.688773+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.00217","created_at":"2026-07-05T12:02:11.688773+00:00"},{"alias_kind":"pith_short_12","alias_value":"UHAXZSIEIUGA","created_at":"2026-07-05T12:02:11.688773+00:00"},{"alias_kind":"pith_short_16","alias_value":"UHAXZSIEIUGAP5YC","created_at":"2026-07-05T12:02:11.688773+00:00"},{"alias_kind":"pith_short_8","alias_value":"UHAXZSIE","created_at":"2026-07-05T12:02:11.688773+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B","json":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B.json","graph_json":"https://pith.science/api/pith-number/UHAXZSIEIUGAP5YCZ2WXTWMK2B/graph.json","events_json":"https://pith.science/api/pith-number/UHAXZSIEIUGAP5YCZ2WXTWMK2B/events.json","paper":"https://pith.science/paper/UHAXZSIE"},"agent_actions":{"view_html":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B","download_json":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B.json","view_paper":"https://pith.science/paper/UHAXZSIE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.00217&json=true","fetch_graph":"https://pith.science/api/pith-number/UHAXZSIEIUGAP5YCZ2WXTWMK2B/graph.json","fetch_events":"https://pith.science/api/pith-number/UHAXZSIEIUGAP5YCZ2WXTWMK2B/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B/action/storage_attestation","attest_author":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B/action/author_attestation","sign_citation":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B/action/citation_signature","submit_replication":"https://pith.science/pith/UHAXZSIEIUGAP5YCZ2WXTWMK2B/action/replication_record"}},"created_at":"2026-07-05T12:02:11.688773+00:00","updated_at":"2026-07-05T12:02:11.688773+00:00"}