{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/marlin-mixed-precision-auto-regressive","title":"MARLIN: Mixed-Precision Auto-Regressive Parallel Inference on Large Language Models","arxiv_id":"2408.11743","date":"2024-08-21","proceeding":null,"authors":["Elias Frantar","Roberto L. Castro","Jiale Chen","Torsten Hoefler","Dan Alistarh"],"abstract":"As inference on Large Language Models (LLMs) emerges as an important workload in machine learning applications, weight quantization has become a standard technique for efficient GPU deployment. Quantization not only reduces model size, but has also been shown to yield substantial speedups for single-user inference, due to reduced memory movement, with low accuracy impact. Yet, it remains open whether speedups are achievable also in \\emph{batched} settings with multiple parallel clients, which are highly relevant for practical serving. It is unclear whether GPU kernels can be designed to remain practically memory-bound, while supporting the substantially increased compute requirements of batched workloads. This paper resolves this question positively by describing the design of Mixed-precision Auto-Regressive LINear kernels, called MARLIN. Concretely, given a model whose weights are compressed via quantization to, e.g., 4 bits per element, MARLIN shows that batchsizes up to 16-32 can be supported with close to maximum ($4\\times$) quantization speedup, and larger batchsizes up to 64-128 with gradually decreasing, but still significant, acceleration. MARLIN accomplishes this via a combination of techniques, such as asynchronous memory access, complex task scheduling and pipelining, and bespoke quantization support. Our experiments show that MARLIN's near-optimal performance on individual LLM layers across different scenarios can also lead to end-to-end LLM inference speedups (of up to $2.8\\times$) when integrated with the popular vLLM serving engine. Finally, MARLIN is extensible to further compression techniques, like NVIDIA 2:4 sparsity, leading to additional speedups.","url_abs":"https://arxiv.org/abs/2408.11743v1","url_pdf":"https://arxiv.org/pdf/2408.11743v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"marlin-mixed-precision-auto-regressive","repo_url":"https://github.com/ist-daslab/marlin","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"marlin-mixed-precision-auto-regressive","repo_url":"https://github.com/ist-daslab/sparse-marlin","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":null,"task_name":"GPU"},{"task_slug":"quantization","task_name":"Quantization"},{"task_slug":"scheduling","task_name":"Scheduling"}],"methods":[{"method_slug":"marlin","method_name":"MARLIN"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2408.11743","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2408.11743"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ist-daslab/sparse-marlin","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ist-daslab/marlin","reach":null}],"summary":{"ran_fixture":3,"ran_honours":1},"by_repo_kind":{"official":{"samples":4,"ran":4,"repositories":2}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"f7a0c2b4b3e78765","entry":"benchmark","repo":"ist-daslab/sparse-marlin","repo_kind":"official","path":"bench_2_4.py","file_url":"https://github.com/ist-daslab/sparse-marlin/blob/HEAD/bench_2_4.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"f7a0c2b4b3e78765"}},{"code_sha256_prefix":"ea86bd4a4cf78d5b","entry":"benchmark_dense","repo":"ist-daslab/marlin","repo_kind":"official","path":"bench.py","file_url":"https://github.com/ist-daslab/marlin/blob/HEAD/bench.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"ea86bd4a4cf78d5b"}},{"code_sha256_prefix":"d74cc21031f4b7c3","entry":"gen_2_4","repo":"ist-daslab/sparse-marlin","repo_kind":"official","path":"bench_2_4.py","file_url":"https://github.com/ist-daslab/sparse-marlin/blob/HEAD/bench_2_4.py","link_basis":"first_harvest_node","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"d74cc21031f4b7c3"}},{"code_sha256_prefix":"7243559c0e01f359","entry":"get_problem","repo":"ist-daslab/sparse-marlin","repo_kind":"official","path":"bench_2_4.py","file_url":"https://github.com/ist-daslab/sparse-marlin/blob/HEAD/bench_2_4.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"7243559c0e01f359"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}