Papers › Transformer in Transformer

Transformer in Transformer

27 Feb 2021NeurIPS 2021 12arXiv:2103.00112archive 2025-07-28

Kai Han, An Xiao, Enhua Wu, Jianyuan Guo, Chunjing Xu, Yunhe Wang

Transformer is a new kind of neural architecture which encodes the input data as powerful features via the attention mechanism. Basically, the visual transformers first divide the input images into several local patches and then calculate both representations and their relationship. Since natural images are of high complexity with abundant detail and color information, the granularity of the patch dividing is not fine enough for excavating features of objects in different scales and locations. In this paper, we point out that the attention inside these local patches are also essential for building visual transformers with high performance and we explore a new architecture, namely, Transformer iN Transformer (TNT). Specifically, we regard the local patches (e.g., 16×16) as "visual sentences" and present to further divide them into smaller patches (e.g., 4×4) as "visual words". The attention of each word will be calculated with other words in the given visual sentence with negligible computational costs. Features of both words and sentences will be aggregated to enhance the representation ability. Experiments on several benchmarks demonstrate the effectiveness of the proposed TNT architecture, e.g., we achieve an 81.5% top-1 accuracy on the ImageNet, which is about 1.7% higher than that of the state-of-the-art visual transformer with similar computational cost. The PyTorch code is available at https://github.com/huawei-noah/CV-Backbones, and the MindSpore code is available at https://gitee.com/mindspore/models/tree/master/research/cv/TNT.

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

Syntology 24 code samples harvested from 4 repositories linked to this paper: 16 ran (of which 12 constructed an object rather than computing a result; 15 with no instrument failure: 1 honoured, 1 violated, 13 with no contract checked; 1 where Syntology's instrument failed) · 8 unverified. Where Syntology's instrument failed: 1 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: official repository: 5 samples from 1 repository, 4 ran (4 with no instrument failure; 0 where Syntology's instrument failed); community (archive-listed): 19 samples from 3 repositories, 12 ran (11 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 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.

huawei-noah/CV-backbones officialmentioned in papermentioned on GitHubtfSyntology: 5 samples, 4 ran (0 instrument-failed) · 1 unverified report
huawei-noah/CV-Backbones officialpytorchSyntology: 5 samples, 4 ran (0 instrument-failed) · 1 unverified report
NZ99/transformer_in_transformer_flax mentioned on GitHubjaxSyntology: 7 samples, 4 ran (0 instrument-failed) · 3 unverified report
Rishit-dagli/Transformer-in-Transformer mentioned on GitHubtfSyntology: 5 samples, 5 ran (1 instrument-failed) · 0 unverified report
lucidrains/transformer-in-transformer mentioned on GitHubpytorchSyntology: 7 samples, 3 ran (0 instrument-failed) · 4 unverified report
2023-MindSpore-1/ms-code-7 mindsporeSyntology: not harvested report
PaddlePaddle/PaddleClas paddleApache-2.0Syntology: not harvested report
mindspore-ai/models mindsporeSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to AdaBin: Improving Binary Neural Networks with Adaptive Binary Sets (arXiv:2208.08084; that paper's own run record: 4 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
mindspore-courses/External-Attention-MindSpore mindsporeSyntology: not harvested report
open-mmlab/mmclassification pytorchApache-2.0Syntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Reversible Vision Transformers (arXiv:2302.04869; that paper's own run record: 24 ran (of which 5 constructed an object rather than computing a result; 17 with no instrument failure: 0 honoured, 0 violated, 17 with no contract checked; 7 where Syntology's instrument failed) · 5 unverified); nothing checks that they implement this paper's method. report
rwightman/pytorch-image-models pytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to AdamP: Slowing Down the Slowdown for Momentum Optimizers on Scale-invariant Weights (arXiv:2006.08217; that paper's own run record: 4 ran (of which 1 constructed an object rather than computing a result; 2 with no instrument failure: 1 honoured, 0 violated, 1 with no contract checked; 2 where Syntology's instrument failed) · 0 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

24 samples harvested: 16 ran (of which 12 constructed an object rather than computing a result; 15 with no instrument failure: 1 honoured, 1 violated, 13 with no contract checked; 1 where Syntology's instrument failed) · 8 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 · honoured contract
1ran · violated contract
1ran · our draft was wrong
13ran
8unverified

Licence: 5 of the 24 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 4 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”. For 2 samples that ran, the graph records a signature this page does not show: for a class the fixture built without calling, the harness called the object with inputs it made up itself, so the recorded signature is the harness's probe, not the fixture's arguments. “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. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

huawei-noah/CV-backbones official repository 5 samples, 4 ran (0 instrument-failed) · 1 unverified
Attention in 344 papers
census: 13 implementations from 12 papers, 10 distinct outputs across 4 buckets, one shared input per bucket
tnt_pytorch/tnt.py ran · metamorphic tier: deterministic no licence file found · pointer only · aa47e20b1de49f21 · report fixture: returned Attention · level recorded 2026-09-22
Block in 252 papers
census: 5 implementations from 5 papers, 3 distinct outputs across 2 buckets, one shared input per bucket
tnt_pytorch/tnt.py ran · metamorphic tier: deterministic no licence file found · pointer only · 0adf903cee3f68c1 · report fixture: returned Block · level recorded 2026-09-22
PatchEmbed in 123 papers
census: 9 implementations from 9 papers, 6 distinct outputs on one shared input
tnt_pytorch/tnt.py ran · metamorphic tier: deterministic no licence file found · pointer only · 3ab31df137c0d6a7 · report fixture: returned PatchEmbed · level recorded 2026-09-22
SE in 13 papers tnt_pytorch/tnt.py ran · metamorphic tier: deterministic no licence file found · pointer only · 3b5e96f71824e712 · report fixture: returned SE · level recorded 2026-09-22
TNT tnt_pytorch/tnt.py unverified no licence file found · pointer only · 5fc54ab39842a479 · report
NZ99/transformer_in_transformer_flax community (archive-listed) 7 samples, 4 ran (0 instrument-failed) · 3 unverified
AddPositionEmbs transformer_in_transformer_flax/tnt.py ran MIT (permissive) · e7fd0617316574c2 · report fixture: returned AddPositionEmbs · level recorded 2026-09-22
MlpBlock in 9 papers transformer_in_transformer_flax/tnt.py ran MIT (permissive) · 0f86fae86e577e92 · report fixture: returned MlpBlock · level recorded 2026-09-22
TNTConfig transformer_in_transformer_flax/tnt.py ran MIT (permissive) · f10d48901b98d6a7 · report fixture: returned TNTConfig · level recorded 2026-09-22
TransformPatches transformer_in_transformer_flax/tnt.py ran MIT (permissive) · 4cb16ca8e61d759a · report fixture: returned TransformPatches · level recorded 2026-09-22
Encoder in 278 papers transformer_in_transformer_flax/tnt.py unverified MIT (permissive) · 6a49d8dc0841e17c · report
TNTBlock transformer_in_transformer_flax/tnt.py unverified MIT (permissive) · 950a07c48a64e5ab · report
TransformerInTransformer transformer_in_transformer_flax/tnt.py unverified MIT (permissive) · 4b8fa19df5d3871a · report
Rishit-dagli/Transformer-in-Transformer community (archive-listed) 5 samples, 5 ran (1 instrument-failed) · 0 unverified
Attention in 344 papers
census: 13 implementations from 12 papers, 10 distinct outputs across 4 buckets, one shared input per bucket
tnt/tnt.py ran Apache-2.0 (permissive) · ae43efde4f9bdb00 · report fixture: returned Attention · level recorded 2026-09-22
FeedForward in 198 papers tnt/tnt.py ran Apache-2.0 (permissive) · fa83b1a389d93eb9 · report fixture: returned FeedForward · level recorded 2026-09-22
PreNorm in 34 papers tnt/tnt.py ran Apache-2.0 (permissive) · bc8ebba1f2a3f25c · report fixture: returned PreNorm · level recorded 2026-09-22
TNT tnt/tnt.py ran Apache-2.0 (permissive) · 3c37926a5a94b20c · report fixture: returned TNT · level recorded 2026-09-22
get_elements_from_nested_list tnt/tnt.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 9a4efbdf001d9928 · report fixture: signature l:list/1/list, new_l:list/1/empty · returned list · level recorded 2026-09-22
lucidrains/transformer-in-transformer community (archive-listed) 7 samples, 3 ran (0 instrument-failed) · 4 unverified
Attention in 344 papers
census: 13 implementations from 12 papers, 10 distinct outputs across 4 buckets, one shared input per bucket
transformer_in_transformer/tnt.py ran · metamorphic tier: invariant MIT (permissive) · 81188c67ea536d0b · report fixture: signature x:Tensor/3/float · returned Tensor · level recorded 2026-09-22
divisible_by in 33 papers transformer_in_transformer/tnt.py ran · violated contract fingerprinted MIT (permissive) · f61d6e6e8177147a · report fixture: signature val:int/0/int, divisor:int/0/int · returned bool · level recorded 2026-09-24
unfold_output_size transformer_in_transformer/tnt.py ran · honoured contract fingerprinted MIT (permissive) · cf4db82cbfb5407c · report fixture: signature image_size:int/0/int, kernel_size:int/0/int, stride:int/0/int, padding:int/0/int · returned int · level recorded 2026-09-22
TNT transformer_in_transformer/tnt.py unverified MIT (permissive) · 19242d4759c5ee91 · report
default in 430 papers transformer_in_transformer/tnt.py unverified MIT (permissive) · b506fc7ff0a3b8a4 · report
divisible_by in 33 papers transformer_in_transformer/tnt.py unverified MIT (permissive) · 65876fc080ac5567 · 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.

4 assemblies on this paper: 4 harvested, 0 generated, listed in 1 groups (siblings lifted in one batch form one group).

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

harvested Assembly for Transformer in Transformer (TNT) (graph method: Transformer in Transformer (TNT)). 6 bricks; individually on a fixture: 5 ran (of which 5 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) · 1 unverified; every one of the 5 samples that ran constructed an object rather than computing a result. The assembly has not been run end-to-end.

one of 4 siblings

  1. TNT entry · level 0 · unverified · 5fc54ab39842a479
  2. Block model · level 1 · ran · 0adf903cee3f68c1
  3. PatchEmbed model · level 1 · ran · 3ab31df137c0d6a7
  4. Attention model · level 1 · ran · aa47e20b1de49f21
  5. Mlp model · level 1 · ran · a1715055832f71fd
  6. SE model · level 1 · ran · 3b5e96f71824e712

harvested Assembly for Transformer in Transformer (TNT) (graph method: Transformer in Transformer (TNT)). 5 bricks; individually on a fixture: 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) · 0 unverified. The assembly has not been run end-to-end.

one of 4 siblings

  1. TNT entry · level 1 · ran · 3c37926a5a94b20c
  2. Attention helper · level 1 · ran · ae43efde4f9bdb00
  3. FeedForward helper · level 1 · ran · fa83b1a389d93eb9
  4. PreNorm helper · level 1 · ran · bc8ebba1f2a3f25c
  5. get_elements_from_nested_list helper · level 1 · ran · our draft was wrong · 9a4efbdf001d9928

harvested Assembly for Transformer in Transformer (TNT) (graph method: Transformer in Transformer (TNT)). 7 bricks; individually on a fixture: 4 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; 0 where Syntology's instrument failed) · 3 unverified; every one of the 4 samples that ran constructed an object rather than computing a result. The assembly has not been run end-to-end.

one of 4 siblings

  1. TransformerInTransformer entry · level 0 · unverified · 4b8fa19df5d3871a
  2. Encoder model · level 0 · unverified · 6a49d8dc0841e17c
  3. TNTConfig helper · level 1 · ran · f10d48901b98d6a7
  4. TransformPatches model · level 1 · ran · 4cb16ca8e61d759a
  5. AddPositionEmbs model · level 1 · ran · e7fd0617316574c2
  6. TNTBlock model · level 0 · unverified · 950a07c48a64e5ab
  7. MlpBlock model · level 1 · ran · 0f86fae86e577e92

harvested Assembly for Transformer in Transformer (TNT) (graph method: Transformer in Transformer (TNT)). 8 bricks; individually on a fixture: 7 ran (of which 2 constructed an object rather than computing a result; 7 with no instrument failure: 1 honoured, 3 violated, 3 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified. The assembly has not been run end-to-end.

one of 4 siblings

  1. TNT entry · level 0 · unverified · 19242d4759c5ee91
  2. Attention helper · level 1 · ran · 81188c67ea536d0b
  3. FeedForward helper · level 1 · ran · bea49fdbe8f8578a
  4. PreNorm helper · level 1 · ran · c40137bc5259e649
  5. default helper · level 1 · ran · violated contract · 60fff7c3c400d7ff
  6. divisible_by helper · level 1 · ran · violated contract · f61d6e6e8177147a
  7. unfold_output_size helper · level 1 · ran · honoured contract · cf4db82cbfb5407c
  8. exists helper · level 1 · ran · violated contract · aa5486a3650902d8

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="2103.00112") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

Fine-Grained Image ClassificationImage ClassificationSentence

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Fine-Grained Image Classification Oxford 102 Flowers TNT-B Accuracy 99.0% #8 of 25 Archive leaderboard report
Fine-Grained Image Classification Oxford 102 Flowers TNT-B PARAMS 65.6M #8 of 25 Archive leaderboard report
Fine-Grained Image Classification Oxford-IIIT Pet Dataset TNT-B Accuracy 95.0% #9 of 15 Archive leaderboard report
Fine-Grained Image Classification Oxford-IIIT Pet Dataset TNT-B PARAMS 65.6M #9 of 15 Archive leaderboard report
Image Classification CIFAR-10 TNT-B Percentage correct 99.1 #15 of 265 Archive leaderboard report
Image Classification CIFAR-100 TNT-B PARAMS 65.6M #21 of 211 Archive leaderboard report
Image Classification CIFAR-100 TNT-B Percentage correct 91.1 #21 of 211 Archive leaderboard report
Image Classification ImageNet TNT-B Number of params 65.6M #380 of 1060 Archive leaderboard report
Image Classification ImageNet TNT-B Top 1 Accuracy 83.9% #380 of 1060 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

Introduced by this paper: TNT

Absolute Position EncodingsAdamAttentionAttention DropoutBPEDeiTDense ConnectionsDropoutFeedforward NetworkLabel SmoothingLayer NormalizationLinear LayerMulti-Head AttentionPosition-Wise Feed-Forward LayerResidual ConnectionSoftmaxTNTTransformer

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