Papers › Analyzing and Improving the Training Dynamics of Diffusion Models

Analyzing and Improving the Training Dynamics of Diffusion Models

5 Dec 2023CVPR 2024 1arXiv:2312.02696archive 2025-07-28

Tero Karras, Miika Aittala, Jaakko Lehtinen, Janne Hellsten, Timo Aila, Samuli Laine

Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling. As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance.

PaperPDFConference PDFCodeCode Syntology ran

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 18 code samples: 15 from 4 repositories linked to this paper, 3 identical to code harvested from another paper's repository. Across all 18: 14 ran (of which 0 constructed an object rather than computing a result; 13 with no instrument failure: 1 honoured, 6 violated, 6 with no contract checked; 1 where Syntology's instrument failed) · 4 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: 2 samples from 1 repository, 2 ran (1 with no instrument failure; 1 where Syntology's instrument failed); community (archive-listed): 13 samples from 3 repositories, 9 ran (9 with no instrument failure; 0 where Syntology's instrument failed); 3 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 3 of its repositories that Syntology harvested for other papers:

Levels written between 2026-09-22 and 2026-09-25 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed. 5 of the 14 samples that ran have no recorded date.

nvlabs/edm2 officialmentioned in papermentioned on GitHubpytorchNOASSERTIONSyntology: 2 samples, 2 ran (1 instrument-failed) · 0 unverified report
FutureXiang/edm2 mentioned on GitHubpytorchSyntology: 1 sample, 1 ran (0 instrument-failed) · 0 unverified report
YichengDWu/tinyedm mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to ScaleLong: Towards More Stable Training of Diffusion Model via Scaling Network Long... (arXiv:2310.13545; that paper's own run record: 13 ran (of which 4 constructed an object rather than computing a result; 5 with no instrument failure: 1 honoured, 0 violated, 4 with no contract checked; 8 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
dopplerchase/cira-diff mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Denoising Diffusion Probabilistic Models (arXiv:2006.11239; that paper's own run record: 186 ran (of which 81 constructed an object rather than computing a result; 142 with no instrument failure: 20 honoured, 8 violated, 114 with no contract checked; 44 where Syntology's instrument failed) · 67 unverified) and Guiding a Diffusion Model with a Bad Version of Itself (arXiv:2406.02507; that paper's own run record: 2 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
lucidrains/ema-pytorch mentioned on GitHubpytorchSyntology: 2 samples, 2 ran (0 instrument-failed) · 0 unverified report
mmathew23/improved_edm mentioned on GitHubpytorchMITSyntology: 10 samples, 6 ran (0 instrument-failed) · 4 unverified report
gohyojun15/ANT_diffusion pytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Perception Prioritized Training of Diffusion Models (arXiv:2204.00227; that paper's own run record: 4 ran (of which 1 constructed an object rather than computing a result; 3 with no instrument failure: 2 honoured, 0 violated, 1 with no contract checked; 1 where Syntology's instrument failed) · 4 unverified), Efficient Diffusion Training via Min-SNR Weighting Strategy (arXiv:2303.09556; that paper's own run record: 3 ran (of which 1 constructed an object rather than computing a result; 3 with no instrument failure: 2 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified) and Addressing Negative Transfer in Diffusion Models (arXiv:2306.00354; that paper's own run record: 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); 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

18 samples harvested: 14 ran (of which 0 constructed an object rather than computing a result; 13 with no instrument failure: 1 honoured, 6 violated, 6 with no contract checked; 1 where Syntology's instrument failed) · 4 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
6ran · violated contract
1ran · our draft was wrong
6ran
4unverified

Licence: 6 of the 18 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. 2 samples whose link to this paper records no repository are listed under one of this paper's own repositories, because identical code was harvested from it for another paper. 3 samples are 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.

nvlabs/edm2 official repository 2 samples, 2 ran (1 instrument-failed) · 0 unverified
edm_sampler in 34 papers generate_images.py identical code Syntology harvested from this repository for another paper; this paper's link records no repository, path or licence of its own ran · our draft was wrong licence of this copy not recorded · a14db9805bf41197 · report fixture: signature net:callable/0/?, noise:Tensor/2/float, labels:Tensor/1/int, gnet:callable/0/?, num_steps:int/0/int, sigma_min:float/0/float, sigma_max:int/0/int, rho:int/0/int, guidance:int/0/int, S_churn:int/0/int, S_min:int/0/int, S_max:float/0/float, S_noise:int/0/int, dtype:dtype/0/?, randn_like:callable/0/? · returned Tensor · level recorded, date not recorded
parse_int_list in 59 papers generate_images.py identical code Syntology harvested from this repository for another paper; this paper's link records no repository, path or licence of its own ran · honoured contract licence of this copy not recorded · cacd4f6ec202d9b4 · report fixture: signature s:str/0/str · returned list · level recorded, date not recorded
FutureXiang/edm2 community (archive-listed) 1 sample, 1 ran (0 instrument-failed) · 0 unverified
get_default_steps sample.py ran · violated contract no licence file found · pointer only · b336b517049dc932 · report fixture: signature model_type:str/0/str, steps:NoneType/0/? · returned int · level recorded 2026-09-22
lucidrains/ema-pytorch community (archive-listed) 2 samples, 2 ran (0 instrument-failed) · 0 unverified
default in 430 papers ema_pytorch/ema_pytorch.py ran · violated contract MIT (permissive) · aa15eaeb5d195280 · report fixture: signature val:NoneType/0/?, d:int/0/int · returned int · level recorded 2026-09-22
first in 17 papers ema_pytorch/post_hoc_ema.py ran · violated contract fingerprinted MIT (permissive) · bf7b2572204d3d13 · report fixture: signature arr:list/1/int · returned int · level recorded 2026-09-22
mmathew23/improved_edm community (archive-listed) 10 samples, 6 ran (0 instrument-failed) · 4 unverified
hide_diffusers_warnings ema_model.py ran MIT (permissive) · 104550381b6dc76a · report fixture: signature func:callable/0/? · returned function · level recorded 2026-09-24
pixel_norm in 13 papers model_a.py ran fingerprinted MIT (permissive) · a0d505b2d6c56221 · report fixture: signature x:Tensor/2/float, eps:float/0/float, dim:int/0/int · returned Tensor · level recorded 2026-09-25
rotate2d augmentation.py ran MIT (permissive) · 892443de8bab233d · report fixture: signature theta:Tensor/0/float · returned Tensor · level recorded 2026-09-24
scale2d augmentation.py ran MIT (permissive) · bc2b81cc2d12e161 · report fixture: signature sx:float/0/float, sy:float/0/float · returned Tensor · level recorded 2026-09-24
translate2d augmentation.py ran MIT (permissive) · c8206975a3365ff0 · report fixture: signature tx:float/0/float, ty:float/0/float · returned Tensor · level recorded 2026-09-24
weight_normalize model.py ran fingerprinted MIT (permissive) · a2beaf5faecf80b8 · report fixture: signature x:Tensor/3/float, eps:float/0/float · returned Tensor · level recorded 2026-09-24
get_down_block in 42 papers model.py unverified MIT (permissive) · 4fa6b80d0288cea1 · report
get_down_block in 42 papers model_a.py unverified MIT (permissive) · eade47d2a50d3683 · report
get_up_block in 43 papers model.py unverified MIT (permissive) · c5a7eb8a1946519a · report
get_up_block in 43 papers model_a.py unverified MIT (permissive) · 3b34295d7d094429 · report
Repository not located 3 samples, 3 ran (0 instrument-failed) · 0 unverified
3 samples are 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.
default in 430 papers repository not located ran · violated contract fingerprinted licence of this copy not recorded · 60fff7c3c400d7ff · report fixture: signature val:int/0/int, d:int/0/int · returned int · level recorded, date not recorded
divisible_by in 33 papers repository not located ran · violated contract fingerprinted licence of this copy not recorded · f912c209e9c899ca · report fixture: signature num:int/0/int, den:int/0/int · returned bool · level recorded, date not recorded
exists in 528 papers
census: 5 implementations from 103 papers, 1 distinct output on one shared input this sample: bucket 1 of 1, arg 1: rank 1, kind int, dtype int64
repository not located ran · violated contract licence of this copy not recorded · aa5486a3650902d8 · report fixture: signature val:Tensor/1/int · returned bool · level recorded, date not recorded

Tasks

Image GenerationPhilosophy

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Generation ImageNet 512x512 EDM2-XXL FID 1.81 #20 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-XXL NFE 126 #20 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-XL FID 1.85 #21 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-XL NFE 126 #21 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-L FID 1.88 #22 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-L NFE 126 #22 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-M FID 2.01 #27 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-M NFE 126 #27 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-S FID 2.23 #31 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-S NFE 126 #31 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-XS FID 2.91 #39 of 52 Archive leaderboard report
Image Generation ImageNet 512x512 EDM2-XS NFE 126 #39 of 52 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

Diffusion

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