Papers › AlphaNet: Improved Training of Supernets with Alpha-Divergence

AlphaNet: Improved Training of Supernets with Alpha-Divergence

16 Feb 2021arXiv:2102.07954archive 2025-07-28

Dilin Wang, Chengyue Gong, Meng Li, Qiang Liu, Vikas Chandra

Weight-sharing neural architecture search (NAS) is an effective technique for automating efficient neural architecture design. Weight-sharing NAS builds a supernet that assembles all the architectures as its sub-networks and jointly trains the supernet with the sub-networks. The success of weight-sharing NAS heavily relies on distilling the knowledge of the supernet to the sub-networks. However, we find that the widely used distillation divergence, i.e., KL divergence, may lead to student sub-networks that over-estimate or under-estimate the uncertainty of the teacher supernet, leading to inferior performance of the sub-networks. In this work, we propose to improve the supernet training with a more generalized alpha-divergence. By adaptively selecting the alpha-divergence, we simultaneously prevent the over-estimation or under-estimation of the uncertainty of the teacher model. We apply the proposed alpha-divergence based supernets training to both slimmable neural networks and weight-sharing NAS, and demonstrate significant improvements. Specifically, our discovered model family, AlphaNet, outperforms prior-art models on a wide range of FLOPs regimes, including BigNAS, Once-for-All networks, and AttentiveNAS. We achieve ImageNet top-1 accuracy of 80.0% with only 444M FLOPs. Our code and pretrained models are available at https://github.com/facebookresearch/AlphaNet.

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 31 code samples: 30 from 2 repositories linked to this paper, 1 identical to code harvested from another paper's repository. Across all 31: 16 ran (of which 0 constructed an object rather than computing a result; 12 with no instrument failure: 4 honoured, 0 violated, 8 with no contract checked; 4 where Syntology's instrument failed) · 15 unverified. Where Syntology's instrument failed: 2 where our contract draft was wrong; 2 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: official repository: 1 sample from 1 repository, 1 ran (0 with no instrument failure; 1 where Syntology's instrument failed); named in the paper: 29 samples from 1 repository, 14 ran (11 with no instrument failure; 3 where Syntology's instrument failed); 1 identical to code first harvested elsewhere. 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 1 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.

facebookresearch/AlphaNet officialmentioned in papermentioned on GitHubpytorchSyntology: 1 sample, 1 ran (1 instrument-failed) · 0 unverified report
facebookresearch/AttentiveNAS mentioned in papermentioned on GitHubpytorchSyntology: 29 samples, 14 ran (3 instrument-failed) · 15 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.

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

31 samples harvested: 16 ran (of which 0 constructed an object rather than computing a result; 12 with no instrument failure: 4 honoured, 0 violated, 8 with no contract checked; 4 where Syntology's instrument failed) · 15 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.

4ran · honoured contract
2ran · our draft was wrong
2ran; the contract check could not run
8ran
15unverified

Licence: 31 of the 31 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. 1 sample is listed last, in a group without a repository. “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. 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.

facebookresearch/AlphaNet official repository 1 sample, 1 ran (1 instrument-failed) · 0 unverified
f_divergence loss_ops.py ran; the contract check could not run (RAISES) fingerprinted licence not identified · pointer only · bd606208d8003d1e · report fixture: signature q_logits:Tensor/2/float, p_logits:Tensor/2/float, alpha:float/0/float, iw_clip:float/0/float · returned tuple · level recorded 2026-09-22
facebookresearch/AttentiveNAS named in the paper 29 samples, 14 ran (3 instrument-failed) · 15 unverified
ConvBnActLayer models/attentive_nas_dynamic_model.py ran fingerprinted licence not identified · pointer only · a7e70ce25e3ac2a4 · report fixture: signature x:Tensor/4/float · returned Tensor · level recorded 2026-09-24
DynamicBatchNorm2d models/attentive_nas_dynamic_model.py ran licence not identified · pointer only · a0efb5eabd43bfb3 · report fixture: returned Tensor · level recorded 2026-09-24
DynamicLinear models/attentive_nas_dynamic_model.py ran · metamorphic tier: invariant fingerprinted licence not identified · pointer only · 86a56f5a600b5419 · report fixture: signature x:Tensor/2/float, out_features:int/0/int · returned Tensor · level recorded 2026-09-22
DynamicLinearLayer models/attentive_nas_dynamic_model.py ran fingerprinted licence not identified · pointer only · 94f0bda6ed5a2e9d · report fixture: signature x:Tensor/2/float · returned Tensor · level recorded 2026-09-24
DynamicPointConv2d models/attentive_nas_dynamic_model.py ran · metamorphic tier: invariant fingerprinted licence not identified · pointer only · df25f4eefb96b548 · report fixture: signature x:Tensor/4/float, out_channel:int/0/int · returned Tensor · level recorded 2026-09-22
IdentityLayer in 7 papers models/attentive_nas_dynamic_model.py ran · metamorphic tier: invariant fingerprinted licence not identified · pointer only · 0257146dc3978be2 · report fixture: signature x:Tensor/2/float · returned Tensor · level recorded 2026-09-22
LinearLayer in 14 papers models/attentive_nas_dynamic_model.py ran · metamorphic tier: invariant licence not identified · pointer only · c3abc01ffd8662f7 · report fixture: signature x:Tensor/2/float · returned Tensor · level recorded 2026-09-22
ShortcutLayer models/attentive_nas_dynamic_model.py ran · metamorphic tier: invariant fingerprinted licence not identified · pointer only · 1c4df2c55cd70330 · report fixture: signature x:Tensor/4/float · returned Tensor · level recorded 2026-09-22
build_activation in 5 papers models/attentive_nas_dynamic_model.py ran · our draft was wrong licence not identified · pointer only · 6150fe6395eb4ee3 · report fixture: signature act_func:str/0/str, inplace:bool/0/bool · returned ReLU · level recorded 2026-09-24
drop_connect in 73 papers
census: 9 implementations from 7 papers, 4 distinct outputs across 2 buckets, one shared input per bucket this sample: bucket 1 of 2, arg 1: rank 4, kind float, dtype float32
models/attentive_nas_dynamic_model.py ran; the contract check could not run (RAISES) licence not identified · pointer only · 6053cb1369b46c73 · report fixture: signature inputs:Tensor/4/float, p:float/0/float, training:bool/0/bool · returned Tensor · level recorded 2026-09-22
get_net_device models/attentive_nas_dynamic_model.py ran · our draft was wrong licence not identified · pointer only · 0e1ee8a546f09486 · report fixture: signature net:callable/0/? · returned device · level recorded 2026-09-22
get_same_padding in 16 papers models/attentive_nas_dynamic_model.py ran · honoured contract fingerprinted licence not identified · pointer only · 71cbea6f98bde1f9 · report fixture: signature kernel_size:int/0/int · returned int · level recorded 2026-09-22
int2list models/attentive_nas_dynamic_model.py ran · honoured contract fingerprinted licence not identified · pointer only · 35ee83149f0affc7 · report fixture: signature val:int/0/int, repeat_time:int/0/int · returned list · level recorded 2026-09-22
sub_filter_start_end models/attentive_nas_dynamic_model.py ran · honoured contract fingerprinted licence not identified · pointer only · 00fdaefcd3c7d435 · report fixture: signature kernel_size:int/0/int, sub_kernel_size:int/0/int · returned tuple · level recorded 2026-09-22
AllReduce models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 0b4bd61096ddb694 · report
AttentiveNasDynamicModel models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 06d08882ecf1c68a · report
AttentiveNasStaticModel models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 8ff71dfd928cae27 · report
DynamicConvBnActLayer models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 39b14fb632110a3f · report
DynamicMBConvLayer models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · c582a9fdd53c2f23 · report
DynamicSE models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 4c386a2f1dbb59bd · report
DynamicSeparableConv2d models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 7949f6e86881267d · report
DynamicShortcutLayer models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · bb02270a794b549e · report
MBInvertedConvLayer models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 1305fda939e9811f · report
MobileInvertedResidualBlock models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · d7620e552fd063af · report
MyModule models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 6669bff300b9aa8f · report
MyNetwork models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · f6b0ce7520ccf192 · report
SELayer in 12 papers models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · a8a91d970372f10e · report
copy_bn models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 551a6373b2274b70 · report
set_layer_from_config models/attentive_nas_dynamic_model.py unverified licence not identified · pointer only · 102325b4dff7f585 · report
Repository not located 1 sample, 1 ran (0 instrument-failed) · 0 unverified
1 sample is identical to code Syntology harvested from another paper's repository; this paper's own copy was not located, so no repository is named and no licence is recorded.
make_divisible in 76 papers repository not located ran · honoured contract fingerprinted licence of this copy not recorded · 3356d3566f053dcc · report fixture: signature v:int/0/int, divisor:int/0/int, min_value:int/0/int · returned int · level recorded 2026-09-22

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 AlphaNet (harvested), lifted in the same batch on 2026-09-09: alternative readings kept side by side; none replaces or supersedes another.

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

one of 2 siblings

  1. f_divergence entry · level 1 · ran; the contract check could not run (RAISES) · bd606208d8003d1e

harvested Assembly for AlphaNet (graph method: AlphaNet). 30 bricks; individually on a fixture: 15 ran (of which 0 constructed an object rather than computing a result; 12 with no instrument failure: 4 honoured, 0 violated, 8 with no contract checked; 3 where Syntology's instrument failed) · 15 unverified. The assembly has not been run end-to-end.

one of 2 siblings

  1. AttentiveNasDynamicModel entry · level 0 · unverified · 06d08882ecf1c68a
  2. AttentiveNasStaticModel helper · level 0 · unverified · 8ff71dfd928cae27
  3. DynamicConvBnActLayer model · level 0 · unverified · 39b14fb632110a3f
  4. DynamicLinearLayer model · level 1 · ran · 94f0bda6ed5a2e9d
  5. DynamicMBConvLayer model · level 0 · unverified · c582a9fdd53c2f23
  6. DynamicShortcutLayer model · level 0 · unverified · bb02270a794b549e
  7. MobileInvertedResidualBlock helper · level 0 · unverified · d7620e552fd063af
  8. MyNetwork helper · level 0 · unverified · f6b0ce7520ccf192
  9. drop_connect helper · level 1 · ran; the contract check could not run (RAISES) · 6053cb1369b46c73
  10. int2list helper · level 1 · ran · honoured contract · 35ee83149f0affc7
  11. make_divisible helper · level 1 · ran · honoured contract · 3356d3566f053dcc
  12. ConvBnActLayer helper · level 1 · ran · a7e70ce25e3ac2a4
  13. DynamicBatchNorm2d model · level 1 · ran · a0efb5eabd43bfb3
  14. DynamicPointConv2d model · level 1 · ran · df25f4eefb96b548
  15. MyModule helper · level 0 · unverified · 6669bff300b9aa8f
  16. build_activation helper · level 1 · ran · our draft was wrong · 6150fe6395eb4ee3
  17. copy_bn helper · level 0 · unverified · 551a6373b2274b70
  18. get_net_device helper · level 1 · ran · our draft was wrong · 0e1ee8a546f09486
  19. DynamicLinear model · level 1 · ran · 86a56f5a600b5419
  20. LinearLayer helper · level 1 · ran · c3abc01ffd8662f7
  21. DynamicSE model · level 0 · unverified · 4c386a2f1dbb59bd
  22. DynamicSeparableConv2d model · level 0 · unverified · 7949f6e86881267d
  23. MBInvertedConvLayer helper · level 0 · unverified · 1305fda939e9811f
  24. SELayer helper · level 0 · unverified · a8a91d970372f10e
  25. ShortcutLayer helper · level 1 · ran · 1c4df2c55cd70330
  26. set_layer_from_config helper · level 0 · unverified · 102325b4dff7f585
  27. get_same_padding helper · level 1 · ran · honoured contract · 71cbea6f98bde1f9
  28. AllReduce helper · level 0 · unverified · 0b4bd61096ddb694
  29. sub_filter_start_end helper · level 1 · ran · honoured contract · 00fdaefcd3c7d435
  30. IdentityLayer helper · level 1 · ran · 0257146dc3978be2

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="2102.07954") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

Image ClassificationNeural Architecture Search

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Classification ImageNet AlphaNet-A6 GFLOPs 0.709 #678 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A6 Top 1 Accuracy 80.8% #678 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A5 GFLOPs 0.491 #707 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A5 Top 1 Accuracy 80.3% #707 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A4 GFLOPs 0.444 #724 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A4 Top 1 Accuracy 80.0% #724 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A3 GFLOPs 0.357 #754 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A3 Top 1 Accuracy 79.4% #754 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A2 GFLOPs 0.317 #774 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A2 Top 1 Accuracy 79.1% #774 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A1 GFLOPs 0.279 #799 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A1 Top 1 Accuracy 78.9% #799 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A0 GFLOPs 0.203 #862 of 1060 Archive leaderboard report
Image Classification ImageNet AlphaNet-A0 Top 1 Accuracy 77.8% #862 of 1060 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A6 Accuracy 80.8 #12 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A6 FLOPs 709M #12 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A6 Top-1 Error Rate 19.2 #12 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A5 (base) Accuracy 80.6 #14 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A5 (base) FLOPs 596M #14 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A5 (base) Top-1 Error Rate 19.4 #14 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A5 (small) Accuracy 80.3 #19 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A5 (small) FLOPs 491M #19 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A5 (small) Top-1 Error Rate 19.7 #19 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A4 Accuracy 80.0 #23 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A4 FLOPs 444M #23 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A4 Top-1 Error Rate 20.0 #23 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A3 Accuracy 79.4 #33 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A3 FLOPs 357M #33 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A3 Top-1 Error Rate 20.6 #33 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A2 Accuracy 79.2 #34 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A2 FLOPs 317M #34 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A2 Top-1 Error Rate 20.8 #34 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A1 Accuracy 79.0 #37 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A1 FLOPs 279M #37 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A1 Top-1 Error Rate 21.0 #37 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A0 Accuracy 77.9 #55 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A0 FLOPs 203M #55 of 135 Archive leaderboard report
Neural Architecture Search ImageNet AlphaNet-A0 Top-1 Error Rate 22.1 #55 of 135 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.

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