Papers › MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models

MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models

26 Sep 2024arXiv:2409.17481archive 2025-07-28

Gongfan Fang, Hongxu Yin, Saurav Muralidharan, Greg Heinrich, Jeff Pool, Jan Kautz, Pavlo Molchanov, Xinchao Wang

Large Language Models (LLMs) are distinguished by their massive parameter counts, which typically result in significant redundancy. This work introduces MaskLLM, a learnable pruning method that establishes Semi-structured (or ``N:M'') Sparsity in LLMs, aimed at reducing computational overhead during inference. Instead of developing a new importance criterion, MaskLLM explicitly models N:M patterns as a learnable distribution through Gumbel Softmax sampling. This approach facilitates end-to-end training on large-scale datasets and offers two notable advantages: 1) High-quality Masks - our method effectively scales to large datasets and learns accurate masks; 2) Transferability - the probabilistic modeling of mask distribution enables the transfer learning of sparsity across domains or tasks. We assessed MaskLLM using 2:4 sparsity on various LLMs, including LLaMA-2, Nemotron-4, and GPT-3, with sizes ranging from 843M to 15B parameters, and our empirical results show substantial improvements over state-of-the-art methods. For instance, leading approaches achieve a perplexity (PPL) of 10 or greater on Wikitext compared to the dense model's 5.12 PPL, but MaskLLM achieves a significantly lower 6.72 PPL solely by learning the masks with frozen weights. Furthermore, MaskLLM's learnable nature allows customized masks for lossless application of 2:4 sparsity to downstream tasks or domains. Code is available at https://github.com/NVlabs/MaskLLM.

PaperPDFCodeCode Syntology ranMethod as code

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 16 code samples harvested from 1 repository linked to this paper: 5 ran (of which 1 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 0 violated, 5 with no contract checked; 0 where Syntology's instrument failed) · 11 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code; here there are none.

By repository: official repository: 16 samples from 1 repository, 5 ran (5 with no instrument failure; 0 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written between 2026-09-22 and 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

nvlabs/maskllm officialmentioned in papermentioned on GitHubpytorchNOASSERTIONSyntology: 16 samples, 5 ran (0 instrument-failed) · 11 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

16 samples harvested: 5 ran (of which 1 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 0 violated, 5 with no contract checked; 0 where Syntology's instrument failed) · 11 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

5ran
11unverified

Licence: 16 of the 16 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from nvlabs/maskllm. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: 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. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

nvlabs/maskllm official repository 16 samples, 5 ran (0 instrument-failed) · 11 unverified
CudaRNGStatesTracker learnable_sparsity/differentiable_mask.py ran licence not identified · pointer only · 3a4d0d8bbbbf7679 · report fixture: returned CudaRNGStatesTracker · level recorded 2026-09-22
get_cuda_rng_tracker learnable_sparsity/differentiable_mask.py ran licence not identified · pointer only · 7073cc1e555d6836 · report fixture: returned CudaRNGStatesTracker · level recorded 2026-09-22
get_expert_parallel_rng_tracker_name learnable_sparsity/differentiable_mask.py ran licence not identified · pointer only · 8d9516b99abbc7b5 · report fixture: returned str · level recorded 2026-09-22
get_llm in 16 papers tool_compute_mask_hf.py ran licence not identified · pointer only · 1b16cd00fbac2ce5 · report fixture: signature model_name:str/0/str, cache_dir:str/0/str · returned Mock · level recorded 2026-09-24
get_wikitext2 in 78 papers eval_llama_ppl.py ran no licence file found · pointer only · dee8be97b2cf4692 · report fixture: returned tuple · level recorded 2026-09-22
DifferentiableMask learnable_sparsity/differentiable_mask.py unverified licence not identified · pointer only · 3feb197bb2f80c7c · report
_ensure_var_is_initialized learnable_sparsity/differentiable_mask.py unverified licence not identified · pointer only · 20d99399acdd4df8 · report
_initialize_affine_weight_gpu learnable_sparsity/differentiable_mask.py unverified licence not identified · pointer only · 6cd59afbca9bb213 · report
_set_cuda_rng_state learnable_sparsity/differentiable_mask.py unverified licence not identified · pointer only · bbfbfd2fc26ae9c4 · report
get_args in 84 papers learnable_sparsity/differentiable_mask.py unverified licence not identified · pointer only · d25230985eb2b130 · report
get_c4 in 41 papers eval_llama_ppl.py unverified no licence file found · pointer only · 7630a157a3b32ab9 · report
get_loaders in 51 papers eval_llama_ppl.py unverified licence not identified · pointer only · a6cf7a5e8f29a402 · report
load_checkpoint_to_model tool_export_to_hf.py unverified licence not identified · pointer only · 91e00b8286e04fd0 · report
parse_args in 355 papers megatron/arguments.py unverified licence not identified · pointer only · f2ce3a79ab6a26c5 · report
print_rank_0 learnable_sparsity/differentiable_mask.py unverified licence not identified · pointer only · 44ee8f559edfd194 · report
set_tensor_model_parallel_attributes learnable_sparsity/differentiable_mask.py unverified licence not identified · pointer only · 5c24809f67be0e0a · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. For a generated assembly, an independent peer is a second implementation generated separately from the same paper text; neither implementation read the authors' code. At level 2 the two agreed numerically on the fixture's inputs; level 3 is recorded when property tests derived from the paper also passed on both, and each level 3 card prints how many its report lists. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it. A generated assembly's label was set when its code was generated for the method, and no model judged it; its card prints it as "by construction" or "not judged by a model".

2 assemblies on this paper: 1 harvested, 1 generated.

generated Assembly for MaskLLM (graph method: MaskLLM), generated from the paper text, not the repository's code. 1 brick; individually on a fixture: 0 ran · 1 unverified. Its end-to-end level was lowered from level 2 to level 0 on 2026-09-10 (level 0: no end-to-end run is now recorded); Syntology's graph records why: “lowered to the entry brick level: a single-routine assembly IS its brick”.

label matches by construction: the code was generated for this method (not judged by a model)

  1. compute_differentiable_mask entry · level 0 · unverified · 90c21487bcb252ff81542caf3830282d160bb7cdc330dd741953c671eea25df1

harvested Assembly for MaskLLM (graph method: MaskLLM). 10 bricks; individually on a fixture: 3 ran (of which 1 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 0 where Syntology's instrument failed) · 7 unverified. The assembly has not been run end-to-end.

  1. DifferentiableMask entry · level 0 · unverified · 3feb197bb2f80c7c
  2. _initialize_affine_weight_gpu helper · level 0 · unverified · 6cd59afbca9bb213
  3. get_args helper · level 0 · unverified · d25230985eb2b130
  4. print_rank_0 helper · level 0 · unverified · 44ee8f559edfd194
  5. get_cuda_rng_tracker helper · level 1 · ran · 7073cc1e555d6836
  6. get_expert_parallel_rng_tracker_name helper · level 1 · ran · 8d9516b99abbc7b5
  7. set_tensor_model_parallel_attributes helper · level 0 · unverified · 5c24809f67be0e0a
  8. _ensure_var_is_initialized helper · level 0 · unverified · 20d99399acdd4df8
  9. _set_cuda_rng_state helper · level 0 · unverified · bbfbfd2fc26ae9c4
  10. CudaRNGStatesTracker helper · level 1 · ran · 3a4d0d8bbbbf7679

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2409.17481") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

Large Language ModelModel CompressionNetwork PruningTransfer Learning

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

AdamAttentionAttention DropoutBPECosine AnnealingDense ConnectionsDropoutGPT-3Gumbel SoftmaxLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingMulti-Head AttentionPruningResidual ConnectionSoftmaxWeight Decay

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections