Papers › Generating Long Sequences with Sparse Transformers

Generating Long Sequences with Sparse Transformers

23 Apr 2019Preprint 2019 4arXiv:1904.10509archive 2025-07-28

Rewon Child, Scott Gray, Alec Radford, Ilya Sutskever

Transformers are powerful sequence models, but require time and memory that grows quadratically with the sequence length. In this paper we introduce sparse factorizations of the attention matrix which reduce this to O(n √(n)). We also introduce a) a variation on architecture and initialization to train deeper networks, b) the recomputation of attention matrices to save memory, and c) fast attention kernels for training. We call networks with these changes Sparse Transformers, and show they can model sequences tens of thousands of timesteps long using hundreds of layers. We use the same architecture to model images, audio, and text from raw bytes, setting a new state of the art for density modeling of Enwik8, CIFAR-10, and ImageNet-64. We generate unconditional samples that demonstrate global coherence and great diversity, and show it is possible in principle to use self-attention to model sequences of length one million or more.

PaperPDFConference PDFCodeCode 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

The archive's official repository (openai/sparse_attention) was not harvested; the 6 samples below come from 2 community repositories.

Syntology 6 code samples harvested from 2 repositories linked to this paper: 5 ran (of which 4 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified. Where Syntology's instrument failed: 1 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: community (archive-listed): 6 samples from 2 repositories, 5 ran (4 with no instrument failure; 1 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.

No sample is linked to this paper from 3 of its repositories that Syntology harvested for other papers:

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

openai/sparse_attention officialmentioned on GitHubtfSyntology: not harvested report
han-shi/SparseBERT mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Attention Is All You Need (arXiv:1706.03762; that paper's own run record: 610 ran (of which 293 constructed an object rather than computing a result; 529 with no instrument failure: 45 honoured, 15 violated, 469 with no contract checked; 81 where Syntology's instrument failed) · 336 unverified), BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (arXiv:1810.04805; that paper's own run record: 300 ran (of which 75 constructed an object rather than computing a result; 234 with no instrument failure: 18 honoured, 5 violated, 211 with no contract checked; 66 where Syntology's instrument failed) · 359 unverified) and SparseBERT: Rethinking the Importance Analysis in Self-attention (arXiv:2102.12871; that paper's own run record: 1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
jonahwinninghoff/Text-Summarization mentioned on GitHubtfSyntology: not harvested report
mistralai/mistral-src mentioned on GitHubpytorchApache-2.0Syntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Mistral 7B (arXiv:2310.06825; that paper's own run record: 10 ran (of which 0 constructed an object rather than computing a result; 10 with no instrument failure: 0 honoured, 0 violated, 10 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
ptillet/torch-blocksparse mentioned on GitHubpytorchSyntology: 2 samples, 2 ran (0 instrument-failed) · 0 unverified report
wilson1yan/VideoGPT mentioned on GitHubpytorchSyntology: 4 samples, 3 ran (1 instrument-failed) · 1 unverified report
MindCode-4/code-11 mindsporeSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Entropy Law: The Story Behind Data Compression and LLM Performance (arXiv:2407.06645; that paper's own run record: 4 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 2 honoured, 1 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. 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.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

6 samples harvested: 5 ran (of which 4 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 1 where Syntology's instrument failed) · 1 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.

1ran · our draft was wrong
4ran
1unverified

Licence: 0 of the 6 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 2 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. “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.

ptillet/torch-blocksparse community (archive-listed) 2 samples, 2 ran (0 instrument-failed) · 0 unverified
DeepSpeedSparseSelfAttention torch_blocksparse/deepspeedsparseselfattention.py ran MIT (permissive) · cc4d4565045cf844 · report fixture: returned DeepSpeedSparseSelfAttention · level recorded 2026-09-22
SparsityConfig torch_blocksparse/deepspeedsparseselfattention.py ran MIT (permissive) · 23a25a6d256367d8 · report fixture: returned SparsityConfig · level recorded 2026-09-22
wilson1yan/VideoGPT community (archive-listed) 4 samples, 3 ran (1 instrument-failed) · 1 unverified
SparseAttention videogpt/attention.py ran · metamorphic tier: deterministic MIT (permissive) · 21937d73418cc532 · report fixture: returned SparseAttention · level recorded 2026-09-22
StridedSparsityConfig videogpt/attention.py ran MIT (permissive) · b5ef9bf53e62eab3 · report fixture: returned StridedSparsityConfig · level recorded 2026-09-22
scaled_dot_product_attention in 43 papers videogpt/attention.py ran · our draft was wrong MIT (permissive) · 9d35276799c65577 · report fixture: signature q:Tensor/3/float, k:Tensor/3/float, v:Tensor/3/float, mask:Tensor/3/float, attn_dropout:float/0/float, training:bool/0/bool · returned Tensor · level recorded 2026-09-22
view_range videogpt/attention.py unverified MIT (permissive) · 9f28796c8276d371 · 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. 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. Siblings are assemblies lifted from this paper in the same batch for the same method: alternative readings kept side by side; none replaces or supersedes another.

2 assemblies on this paper: 2 harvested, 0 generated, listed in 1 groups (siblings lifted in one batch form one group).

2 sibling assemblies for Sparse Transformers (harvested), lifted in the same batch on 2026-09-09: alternative readings kept side by side; none replaces or supersedes another.

harvested Assembly for Sparse Transformers (graph method: Sparse Transformers). 4 bricks; individually on a fixture: 3 ran (of which 2 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified. The assembly has not been run end-to-end.

one of 2 siblings

  1. SparseAttention entry · level 1 · ran · 21937d73418cc532
  2. StridedSparsityConfig helper · level 1 · ran · b5ef9bf53e62eab3
  3. scaled_dot_product_attention helper · level 1 · ran · our draft was wrong · 9d35276799c65577
  4. view_range helper · level 0 · unverified · 9f28796c8276d371

harvested Assembly for Sparse Transformers (graph method: Sparse Transformers). 2 bricks; individually on a fixture: 2 ran (of which 2 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified; every one of the 2 samples that ran constructed an object rather than computing a result. The assembly has not been run end-to-end.

one of 2 siblings

  1. DeepSpeedSparseSelfAttention entry · level 1 · ran · cc4d4565045cf844
  2. SparsityConfig helper · level 1 · ran · 23a25a6d256367d8

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="1904.10509") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

DiversityImage GenerationLanguage ModellingOpen-Domain Question AnsweringQuestion Answering

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Audio Generation Classical music, 5 seconds at 12 kHz Sparse Transformer 152M (strided) Bits per byte 1.97 #1 of 2 Archive leaderboard report
Image Generation ImageNet 64x64 Sparse Transformer 59M (strided) Bits per dim 3.44 #43 of 65 Archive leaderboard report
Language Modelling enwik8 Sparse Transformer (30 layers, fixed attn) Bit per Character (BPC) 0.99 #14 of 42 Archive leaderboard report
Language Modelling enwik8 Sparse Transformer (30 layers, fixed attn) Number of params 95M #14 of 42 Archive leaderboard report
Open-Domain Question Answering SearchQA Sparse Attention EM 64.7 #4 of 14 Archive leaderboard report
Question Answering Natural Questions (long) Sparse Attention F1 74.5 #4 of 13 Archive leaderboard report
Question Answering Quasart-T Sparse Attention EM 52.1 #3 of 7 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AdamAttentionAttention DropoutCosine AnnealingDense ConnectionsDropoutLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingMulti-Head AttentionResidual ConnectionSoftmaxSparse TransformerWeight 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