{"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/u-m-p-the-unit-scaled-maximal-update","title":"u-$μ$P: The Unit-Scaled Maximal Update Parametrization","arxiv_id":"2407.17465","date":"2024-07-24","proceeding":null,"authors":["Charlie Blake","Constantin Eichenberg","Josef Dean","Lukas Balles","Luke Y. Prince","Björn Deiseroth","Andres Felipe Cruz-Salinas","Carlo Luschi","Samuel Weinbach","Douglas Orr"],"abstract":"The Maximal Update Parametrization ($\\mu$P) aims to make the optimal hyperparameters (HPs) of a model independent of its size, allowing them to be swept using a cheap proxy model rather than the full-size target model. We present a new scheme, u-$\\mu$P, which improves upon $\\mu$P by combining it with Unit Scaling, a method for designing models that makes them easy to train in low-precision. The two techniques have a natural affinity: $\\mu$P ensures that the scale of activations is independent of model size, and Unit Scaling ensures that activations, weights and gradients begin training with a scale of one. This synthesis opens the door to a simpler scheme, whose default values are near-optimal. This in turn facilitates a more efficient sweeping strategy, with u-$\\mu$P models reaching a loss that is equal to or lower than comparable $\\mu$P models and working out-of-the-box in FP8.","url_abs":"https://arxiv.org/abs/2407.17465v3","url_pdf":"https://arxiv.org/pdf/2407.17465v3.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":"u-m-p-the-unit-scaled-maximal-update","repo_url":"https://github.com/graphcore-research/unit-scaling","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2407.17465","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2407.17465"}},"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/graphcore-research/unit-scaling","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran":12,"unverified":1},"by_repo_kind":{"official":{"samples":13,"ran":12,"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":"8d7406030299bc46","entry":"Parameter","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/parameter.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/parameter.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":"8d7406030299bc46"}},{"code_sha256_prefix":"070a400b0ece5434","entry":"format_to_tuple","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/formats.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/formats.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":"070a400b0ece5434"}},{"code_sha256_prefix":"de8aa1c681baa450","entry":"generate__all__","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/_internal_utils.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/_internal_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":"de8aa1c681baa450"}},{"code_sha256_prefix":"d19cc5ff648afebd","entry":"has_parameter_data","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/parameter.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/parameter.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":"d19cc5ff648afebd"}},{"code_sha256_prefix":"e0c95afd064e25ed","entry":"logarithmic_interpolation","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/core/functional.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/core/functional.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":"e0c95afd064e25ed"}},{"code_sha256_prefix":"a1e2fc6500a07813","entry":"mult_docstring","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/docs.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/docs.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":"a1e2fc6500a07813"}},{"code_sha256_prefix":"7f174b2d78b36837","entry":"rms","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/core/functional.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/core/functional.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":"7f174b2d78b36837"}},{"code_sha256_prefix":"af14bc3abf24163b","entry":"scale_bwd","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/scale.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/scale.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":"af14bc3abf24163b"}},{"code_sha256_prefix":"99e5d990030cdf9a","entry":"scale_fwd","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/scale.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/scale.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":"99e5d990030cdf9a"}},{"code_sha256_prefix":"311ddf75624e8833","entry":"to_grad_input_scale","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/constraints.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/constraints.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":"311ddf75624e8833"}},{"code_sha256_prefix":"e57d5ac2fdb91999","entry":"to_left_grad_scale","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/constraints.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/constraints.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":"e57d5ac2fdb91999"}},{"code_sha256_prefix":"07ba3009b696a3c2","entry":"tuple_to_format","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/formats.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/formats.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":"07ba3009b696a3c2"}},{"code_sha256_prefix":"873f0a22bcd0444f","entry":"to_output_scale","repo":"graphcore-research/unit-scaling","repo_kind":"official","path":"unit_scaling/constraints.py","file_url":"https://github.com/graphcore-research/unit-scaling/blob/HEAD/unit_scaling/constraints.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":"873f0a22bcd0444f"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}