{"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/pv-tuning-beyond-straight-through-estimation","title":"PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compression","arxiv_id":"2405.14852","date":"2024-05-23","proceeding":null,"authors":["Vladimir Malinovskii","Denis Mazur","Ivan Ilin","Denis Kuznedelev","Konstantin Burlachenko","Kai Yi","Dan Alistarh","Peter Richtarik"],"abstract":"There has been significant interest in \"extreme\" compression of large language models (LLMs), i.e., to 1-2 bits per parameter, which allows such models to be executed efficiently on resource-constrained devices. Existing work focused on improved one-shot quantization techniques and weight representations; yet, purely post-training approaches are reaching diminishing returns in terms of the accuracy-vs-bit-width trade-off. State-of-the-art quantization methods such as QuIP# and AQLM include fine-tuning (part of) the compressed parameters over a limited amount of calibration data; however, such fine-tuning techniques over compressed weights often make exclusive use of straight-through estimators (STE), whose performance is not well-understood in this setting. In this work, we question the use of STE for extreme LLM compression, showing that it can be sub-optimal, and perform a systematic study of quantization-aware fine-tuning strategies for LLMs. We propose PV-Tuning - a representation-agnostic framework that generalizes and improves upon existing fine-tuning strategies, and provides convergence guarantees in restricted cases. On the practical side, when used for 1-2 bit vector quantization, PV-Tuning outperforms prior techniques for highly-performant models such as Llama and Mistral. Using PV-Tuning, we achieve the first Pareto-optimal quantization for Llama 2 family models at 2 bits per parameter.","url_abs":"https://arxiv.org/abs/2405.14852v2","url_pdf":"https://arxiv.org/pdf/2405.14852v2.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":"pv-tuning-beyond-straight-through-estimation","repo_url":"https://github.com/vahe1994/aqlm","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"quantization","task_name":"Quantization"}],"methods":[{"method_slug":"llama","method_name":"LLaMA"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2405.14852","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2405.14852"}},"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/vahe1994/aqlm","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran":5,"unverified":5},"by_repo_kind":{"official":{"samples":10,"ran":5,"repositories":1}},"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":"286ca2263c661514","entry":"get_int_dtype","repo":"vahe1994/aqlm","repo_kind":"official","path":"inference_lib/src/aqlm/utils.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/inference_lib/src/aqlm/utils.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"286ca2263c661514"}},{"code_sha256_prefix":"18920618b7b5ee81","entry":"get_num_layers","repo":"vahe1994/aqlm","repo_kind":"official","path":"convert_to_hf.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/convert_to_hf.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"18920618b7b5ee81"}},{"code_sha256_prefix":"9782bd07382e951c","entry":"next_power_of_2","repo":"vahe1994/aqlm","repo_kind":"official","path":"inference_lib/src/aqlm/inference_kernels/triton_kernel.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/inference_lib/src/aqlm/inference_kernels/triton_kernel.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"9782bd07382e951c"}},{"code_sha256_prefix":"0e1845ecac45aea3","entry":"pack_int_data","repo":"vahe1994/aqlm","repo_kind":"official","path":"inference_lib/src/aqlm/utils.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/inference_lib/src/aqlm/utils.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"0e1845ecac45aea3"}},{"code_sha256_prefix":"a1a49cbe29b10116","entry":"unpack_int_data","repo":"vahe1994/aqlm","repo_kind":"official","path":"inference_lib/src/aqlm/utils.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/inference_lib/src/aqlm/utils.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"a1a49cbe29b10116"}},{"code_sha256_prefix":"8443c635b944859f","entry":"get_backward_pass_kernel","repo":"vahe1994/aqlm","repo_kind":"official","path":"inference_lib/src/aqlm/inference_kernels/kernel_selector.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/inference_lib/src/aqlm/inference_kernels/kernel_selector.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"8443c635b944859f"}},{"code_sha256_prefix":"4538301eff106753","entry":"get_forward_pass_kernel","repo":"vahe1994/aqlm","repo_kind":"official","path":"inference_lib/src/aqlm/inference_kernels/kernel_selector.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/inference_lib/src/aqlm/inference_kernels/kernel_selector.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"4538301eff106753"}},{"code_sha256_prefix":"db8f26f618d4290e","entry":"numba_gemm_lut","repo":"vahe1994/aqlm","repo_kind":"official","path":"inference_lib/src/aqlm/inference_kernels/numba_kernel.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/inference_lib/src/aqlm/inference_kernels/numba_kernel.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"db8f26f618d4290e"}},{"code_sha256_prefix":"44925cf3831bfe79","entry":"rgetattr","repo":"vahe1994/aqlm","repo_kind":"official","path":"convert_legacy_model_format.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/convert_legacy_model_format.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"44925cf3831bfe79"}},{"code_sha256_prefix":"f9de4cdd8a624801","entry":"rsetattr","repo":"vahe1994/aqlm","repo_kind":"official","path":"convert_legacy_model_format.py","file_url":"https://github.com/vahe1994/aqlm/blob/HEAD/convert_legacy_model_format.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"f9de4cdd8a624801"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}