{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:P4GUNJ2ZIR5ZBHVODMM6GOWR3C","short_pith_number":"pith:P4GUNJ2Z","schema_version":"1.0","canonical_sha256":"7f0d46a759447b909eae1b19e33ad1d8866f4fbaa12d466eddc6814c83ed3ab3","source":{"kind":"arxiv","id":"2405.03594","version":1},"attestation_state":"computed","paper":{"title":"Enabling High-Sparsity Foundational Llama Models with Efficient Pretraining and Deployment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Abhay Gupta, Abhinav Agarwalla, Alexandre Marques, Dan Alistarh, Eldar Kurtic, Kevin Leong, Mahmoud Salem, Mark Kurtz, Michael Goin, Sean Lie, Shubhra Pandit, Tuan Nguyen","submitted_at":"2024-05-06T16:03:32Z","abstract_excerpt":"Large language models (LLMs) have revolutionized Natural Language Processing (NLP), but their size creates computational bottlenecks. We introduce a novel approach to create accurate, sparse foundational versions of performant LLMs that achieve full accuracy recovery for fine-tuning tasks at up to 70% sparsity. We achieve this for the LLaMA-2 7B model by combining the SparseGPT one-shot pruning method and sparse pretraining of those models on a subset of the SlimPajama dataset mixed with a Python subset of The Stack dataset. We exhibit training acceleration due to sparsity on Cerebras CS-3 chi"},"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":"2405.03594","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-06T16:03:32Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"93ad91e60e417f78d87d738cb280baf48b2bc6835befc7ed5312b8dba6851fec","abstract_canon_sha256":"50f37e9d3db00d86cd07b4cf00b81c2534d66b83b28026a6d8e96608985305f5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:16:04.961251Z","signature_b64":"5q4mAm2zh8suOqgI+6IXeHr3bPbrIfVFLRa27VMNMfEwZokThAggfZx82Y3SIMMXz925g++0iDDmtpq8ztQLBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7f0d46a759447b909eae1b19e33ad1d8866f4fbaa12d466eddc6814c83ed3ab3","last_reissued_at":"2026-07-05T08:16:04.960763Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:16:04.960763Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Enabling High-Sparsity Foundational Llama Models with Efficient Pretraining and Deployment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Abhay Gupta, Abhinav Agarwalla, Alexandre Marques, Dan Alistarh, Eldar Kurtic, Kevin Leong, Mahmoud Salem, Mark Kurtz, Michael Goin, Sean Lie, Shubhra Pandit, Tuan Nguyen","submitted_at":"2024-05-06T16:03:32Z","abstract_excerpt":"Large language models (LLMs) have revolutionized Natural Language Processing (NLP), but their size creates computational bottlenecks. We introduce a novel approach to create accurate, sparse foundational versions of performant LLMs that achieve full accuracy recovery for fine-tuning tasks at up to 70% sparsity. We achieve this for the LLaMA-2 7B model by combining the SparseGPT one-shot pruning method and sparse pretraining of those models on a subset of the SlimPajama dataset mixed with a Python subset of The Stack dataset. We exhibit training acceleration due to sparsity on Cerebras CS-3 chi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.03594","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/2405.03594/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":"2405.03594","created_at":"2026-07-05T08:16:04.960820+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.03594v1","created_at":"2026-07-05T08:16:04.960820+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.03594","created_at":"2026-07-05T08:16:04.960820+00:00"},{"alias_kind":"pith_short_12","alias_value":"P4GUNJ2ZIR5Z","created_at":"2026-07-05T08:16:04.960820+00:00"},{"alias_kind":"pith_short_16","alias_value":"P4GUNJ2ZIR5ZBHVO","created_at":"2026-07-05T08:16:04.960820+00:00"},{"alias_kind":"pith_short_8","alias_value":"P4GUNJ2Z","created_at":"2026-07-05T08:16:04.960820+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01155","citing_title":"When Data Is Scarce: Scaling Sparse Language Models with Repeated Training","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C","json":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C.json","graph_json":"https://pith.science/api/pith-number/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/graph.json","events_json":"https://pith.science/api/pith-number/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/events.json","paper":"https://pith.science/paper/P4GUNJ2Z"},"agent_actions":{"view_html":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C","download_json":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C.json","view_paper":"https://pith.science/paper/P4GUNJ2Z","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.03594&json=true","fetch_graph":"https://pith.science/api/pith-number/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/graph.json","fetch_events":"https://pith.science/api/pith-number/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/action/timestamp_anchor","attest_storage":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/action/storage_attestation","attest_author":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/action/author_attestation","sign_citation":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/action/citation_signature","submit_replication":"https://pith.science/pith/P4GUNJ2ZIR5ZBHVODMM6GOWR3C/action/replication_record"}},"created_at":"2026-07-05T08:16:04.960820+00:00","updated_at":"2026-07-05T08:16:04.960820+00:00"}