TCAQ-DM: Timestep-Channel Adaptive Quantization for Diffusion Models

Haocheng Huang; Jiaxin Chen; Jinyang Guo; Ruiyi Zhan; Yunhong Wang

REVIEW 3 major objections 6 minor 52 references

TCAQ-DM: Timestep-Channel Adaptive Quantization for Diffusion Models

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A post-training quantization method, TCAQ-DM, claims 4-bit diffusion models can generate usable images, cutting CIFAR-10 FID from over 236 to 6.38.

desk verdict Strong W4A4 PTQ results for diffusion models, but the headline tables may be comparing 8-bit softmax against lower-bit softmax baselines without disclosing it; still worth a serious referee. read the letter →

arxiv 2412.16700 v1 pith:HJY7O7SW submitted 2024-12-21 cs.CV

Haocheng Huang , Jiaxin Chen , Jinyang Guo , Ruiyi Zhan , Yunhong Wang This is my paper · ORCID

classification cs.CV

keywords post-trainingquantizationdiffusionmodelslow-bitactivationreparameterizationadaptivequantizercalibrationalignmentFIDimagegeneration

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Diffusion models produce high-quality images but are expensive to run, and post-training quantization can cut that cost with little retraining. The paper argues that prior quantization methods fail on diffusion models because activation ranges swing wildly across channels and timesteps, and because the data used during calibration does not match the data seen during iterative inference. To fix this, it proposes three linked modules: a timestep-channel reparameterization that evens out activation ranges, a dynamically adaptive quantizer that picks log or uniform quantization per timestep for softmax outputs, and a progressively aligned reconstruction that recalibrates on samples drawn from the partially quantized model. On CIFAR-10, the method reports FID 4.40 at 6-bit weights and activations against 4.14 for the full-precision model, and FID 6.38 at 4-bit weights and activations where prior methods collapse to FIDs above 200.

What carries the argument

The load-bearing object is a per-channel, per-timestep scaling vector derived from the pooled maximum activation of each channel across timesteps, applied as a reparameterization that shifts value range from activations into weights, plus a per-layer, per-timestep likelihood-ratio test that selects between a $\log_2$ quantizer and a uniform quantizer for post-Softmax activations. The third piece is an iterative recalibration loop, described in Algorithm 1, in which the calibration set is resampled from the partially quantized model and used to refine quantization parameters over successive rounds. Together they let a single set of quantization parameters handle distributions that change shape as denoising proceeds.

What would settle it

Run the PAR procedure for 4-bit weights and activations on CIFAR-10 with two variants: resampling from the quantized model (as proposed) and resampling from the full-precision model at every round. If the full-precision-resampled control matches or beats the proposed PAR FID of 6.38, the alignment mechanism is not responsible for the gain and the central claim about input mismatch is undercut.

Watch

Extended reading notes

Core claim

The central claim is that a post-training quantization pipeline can quantize diffusion models to very low bit-widths without catastrophic quality loss, provided it adapts to three properties of diffusion computation: the joint fluctuation of activation ranges over timesteps and channels, the timestep-dependent shape of post-Softmax distributions, and the input mismatch between one-shot calibration and iterative inference. The authors introduce the timestep-channel joint reparameterization (TCR) module, which rescales activations channel-wise using a timestep-averaged weight so that outlier channels are tamed; the dynamically adaptive quantizer (DAQ), which uses a maximum-likelihood power-law fit to decide per timestep whether a $\log_2$ quantizer or a uniform quantizer better matches the post-Softmax distribution; and the progressively aligned reconstruction (PAR) strategy, which resamples the calibration set from the quantized model during reconstruction so the inputs match inference-time data flow. Supported by ablations, each module is reported to contribute to the final FID, with the full method producing usable images at 4-bit weights and activations where baseline PTQ methods produce FIDs of hundreds.

Load-bearing premise

The method assumes that sampling a fresh calibration set from the partially quantized model yields inputs representative of what the final quantized model will see during iterative inference, so if the early quantized outputs are badly distorted the recalibration could reinforce the distortion rather than correct it.

Editorial extensions

If this is right

At 6-bit weights and activations, quantized DDIM on CIFAR-10 reaches FID 4.40, nearly matching the full-precision 4.14, so low-bit diffusion can run with little perceivable quality loss.
At 4-bit weights and activations, the method produces usable images (FID 6.38 on CIFAR-10) where prior PTQ methods collapse to FID over 200, opening W4A4 as a practical operating point.
The same three modules transfer across unconditional and conditional generation: results are reported on CIFAR-10, LSUN-Bedrooms, LSUN-Churches, and ImageNet with DDIM and LDM-4.
DAQ's per-timestep quantizer selection yields stable gains across softmax bit-widths from 8 down to 4 bits, whereas fixed log or uniform quantizers degrade sharply.
Because the method is post-training with a small calibration set, it avoids retraining and can be deployed on a single RTX 4090, per the paper's implementation details.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The PAR loop assumes the quantized model's own outputs are a good proxy for inference inputs; one could test this by comparing PAR against a control that resamples from the full-precision model at every round, and if the control performs equally well, the alignment effect is not actually driving the gain.
The TCR reparameterization is not limited to diffusion: any temporally iterated network with channel-wise activation outliers, such as video or recurrent models, could adopt the same timestep-averaged rescaling.
The DAQ likelihood-ratio choice between log and uniform quantizers could be extended to more than two quantizer families, or applied to other heavy-tailed activation layers beyond Softmax, such as certain normalization outputs.
The reported W4A4 gains hinge on keeping the input and output layers and the post-Softmax layer in higher precision; a fully 4-bit pipeline would need further work.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The short version: this is a genuinely useful PTQ-for-diffusion paper with strong W4A4 numbers, but the headline results may be quoted with an 8-bit post-Softmax layer while some baselines are quoted at lower softmax precision. If that's true, the claimed advantage over prior art is partly apples-to-oranges. The fix is easy—report S in every table—and the underlying method still looks sensible.

What's new: they combine timestep-channel reparameterization (TCR), per-timestep quantizer selection for post-Softmax layers (DAQ), and progressive recalibration from the quantized model (PAR). The combination is genuinely new, and the W4A4 results are impressive: 6.38 FID on CIFAR-10 versus collapsed baselines in the hundreds is a real advance, even if the softmax is at 8-bit. The algebra for TCR is correct—it's a valid linear reparameterization that preserves the output—and the ablations are consistent with the story. The comparison across three datasets and two model families is broader than most papers in this area.

The soft spots: first, the post-Softmax bit-width ambiguity is load-bearing. The implementation details say "all experiments are conducted with an 8-bit post-Softmax layer unless being specifically claimed." Table 1 has no S column, so its W6A6 and W4A4 rows are presumably W6A6S8 and W4A4S8. Baselines re-run from open code may have used lower softmax precision. That would make the W4A4 comparison in Table 1 unfair. Table 2 explicitly reports S8/S32, so the authors know the difference; they just didn't apply it consistently to Table 1. This is a reporting flaw, but it changes the meaning of the headline numbers.

Second, there are no error bars, no code or checkpoints released, and several hyperparameters (TCR clamp range, DAQ likelihood threshold, PAR rounds) are tuned on the benchmarks themselves. The PAR resampling mechanism—recalibrating with samples from the partially quantized model—could in principle amplify distortion, though the ablation shows it helps. A control with full-precision recalibration would strengthen that section. These are fixable but currently make exact replication hard.

Despite the reporting issues, the method is coherent and the W4A4 results, even with S8 softmax, are worth taking seriously. I'd send this to peer review. The authors need to disclose softmax bit-width in every table, release code, and run the PAR control. If the numbers hold under those conditions, it's a solid acceptance.

Referee Report

3 major / 6 minor

Summary. This paper proposes TCAQ-DM, a post-training quantization (PTQ) method for diffusion models. The method combines three components: (i) a timestep-channel joint reparameterization (TCR) module that rescales convolution activations along channels with timestep-aware weights; (ii) a dynamically adaptive quantizer (DAQ) that selects between log2 and uniform quantizers for post-Softmax layers based on a likelihood-ratio criterion; and (iii) a progressively aligned reconstruction (PAR) strategy that re-samples the calibration set from the partially quantized model during reconstruction. The method is evaluated on CIFAR-10 with DDIM, LSUN-Bedrooms/LSUN-Churches with LDM-4, and ImageNet with LDM-4, reporting FID, sFID, and IS. The headline results are W6A6 FID 4.40 versus 4.14 for the full-precision model on CIFAR-10, and W4A4 FID 6.38 versus 236.63 for the best prior method.

Significance. If the reported bit-widths are made fully transparent, this is a practically significant contribution: the W4A4 numbers are dramatically better than prior PTQ methods for diffusion models, and the TCR reparameterization is mathematically sound because channel-wise rescaling of inputs and weights preserves the convolution output. The ablation study in Table 5 is consistent with each component adding value, and the evaluation spans multiple datasets and two diffusion-model families. The main reservation is that the paper's default 8-bit post-Softmax setting makes the headline W4A4/W6A6 labels incomplete unless the tables state the post-Softmax width for every entry.

major comments (3)

[Implementation Details / Tables 1, 3, 4] The Implementation Details state that all experiments use an 8-bit post-Softmax layer unless specifically claimed, but Tables 1, 3, and 4 do not include a post-Softmax bit-width column or a caption note. As a result, the labels "W6A6" and "W4A4" for the proposed method in these tables are incomplete: the post-Softmax activations are S8, not 6-bit or 4-bit. This is not purely cosmetic, because the comparisons against PTQ4DM*, Q-Diffusion*, and TFMQ-DM* in W4A4 may pair an S8 model with baselines whose post-Softmax width differs. Please add an S column (or explicit caption statements) to Tables 1, 3, and 4, and for every baseline report the exact post-Softmax bit-width used; if some baselines use S32 or S4, the comparison should be re-labeled so that the reader can see whether the comparison is same-precision.
[Progressively Aligned Reconstruction / Table 5] The PAR module resamples the calibration set from the partially quantized model (Algorithm 1, lines 5-6). The ablation in Table 5 attributes a 2.71 FID improvement at W4A4 (from 9.09 to 6.38) to PAR, but the design does not isolate the effect of aligned sampling from the effect of additional reconstruction iterations, since each PAR round performs 10,000 extra iterations. Without a control that runs the same number of extra iterations while drawing calibration samples from the full-precision model, or that keeps the calibration set fixed, the specific mechanism behind the PAR gain is not established. Please add such a control experiment.
[Dynamically Adaptive Quantizer / Eq. (8)] The selection variable Rg is introduced as a "ratio" of likelihood estimation results but is compared against zero. A likelihood ratio is non-negative, so either Rg is a log-likelihood ratio or the definition is incomplete. The text should give the exact formula for Rg, specify the candidate distributions (power law, log-normal, exponential) and how the MLE fit is performed per layer and timestep, and clarify the decision rule in Eq. (8). Without this information, the DAQ module is not reproducible.

minor comments (6)

[Methodology, Eq. (3) vs Eq. (5)] The subscript in Eq. (3) uses j while Eq. (5) uses d for the same channel index; please unify the notation.
[Methodology vs Algorithm 1] The text refers to "the basic reconstruction with BRECQ" while Algorithm 1 says "perform the basic Adaround operation"; please make the reconstruction procedure terminology consistent.
[Table 2 caption] The caption says "except for the post-Softmax quantization bit-width" but the table lists several S values for both our method and the baselines; please spell out which setting applies to which method.
[Experimental Results] No error bars or multiple seeds are reported; several high-bit comparisons have small FID differences (for example, 3.21 vs 3.14 in Table 2), so reporting variance or the number of evaluation seeds would help the reader judge significance.
[Abstract] In the abstract, "fail to tackle with the large variations" should be "fail to handle the large variations".
[Supplementary Material, Figure C] The reference "Fig. Ca (b) and (c)" appears to be a typo; it should be "Fig. C(b) and C(c)".

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: TCAQ-DM's claims rest on external FID/IS benchmarks and component ablations, not on self-referential fits or self-citations.

full rationale

The paper's central claims are empirical and externally benchmarked. TCAQ-DM includes three components: TCR reparameterizes activations with a data-derived scaling vector while preserving the layer output; DAQ selects between log2 and uniform quantizers per timestep based on a power-law likelihood ratio computed offline from collected activations; PAR iteratively resamples the calibration set from the partially quantized model and reruns AdaRound reconstruction. None of these components define the reported FID/IS values in terms of themselves. The reported headline numbers are generated by the resulting quantized model and compared against full-precision FID and re-run baselines; ablations (Table 5) show incremental FID changes attributable to each module. The method cites prior work (BRECQ, RepQ-ViT, FQ-ViT, TFMQ-DM) for pieces, but the core mechanism is not justified by a uniqueness theorem or a self-citation chain. The one transparency caveat—the implementation note 'All experiments are conducted with an 8-bit post-Softmax layer unless being specifically claimed'—may make the Table 1 W4A4/W6A6 labels ambiguous relative to baselines, but this is an experimental-reporting concern, not circular derivation. No fitted parameter is renamed as a prediction and no equation reduces to its own input by construction.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The method introduces no physical quantities or entities. Its load-bearing assumptions are modeling assumptions about activation distributions and about using the quantized model's own samples for calibration. The listed free parameters are design choices tuned on the evaluation benchmark.

free parameters (4)

TCR clamp range R_tru (r_c) = 5 for W4A8 on CIFAR-10; 100/50/10/3 tried in Table A; values for W4A4 and other settings not reported
Truncates the channel reparameterization scaling vector to avoid enlarging weights too much under low-bit weight quantization; selected by sweeping FID on the same CIFAR-10 benchmark where headline results are reported.
DAQ likelihood-ratio threshold = 0 (use log2 quantizer when Rg > 0)
Binary decision rule in Eq. (8); the threshold and the exact definition of Rg are not justified or ablated.
PAR rounds and extra iterations = 2 additional progressive rounds and 10,000 iterations per round
Selected by ablation on CIFAR-10 DDIM (supplementary Fig. C); no validation split is specified.
Timestep group count G = not reported
The text says the timesteps are uniformly split into groups for TCR, but no group number or grouping formula appears in Section 3; this affects storage and quantization granularity.

assumptions (4)

standard math Convolution with input channels scaled by 1/r_d and weights scaled by r_d has exactly the same output.
Eqs. (3) and (5); exact in full precision for a linear convolution, but weight rounding at low bit-widths breaks it, which is why a clamp range is introduced.
domain assumption A single timestep-independent channel scaling vector r^s can balance activation ranges across all timesteps without loss of information.
Eq. (5) aggregates r_t with max-weighted averaging across timesteps; if activation ranges in different timesteps differ strongly, one vector cannot normalize them all.
domain assumption Post-Softmax activations in diffusion models are either power-law-like or concentrated, and the MLE likelihood ratio identifies which quantizer is better.
Section 'Dynamically Adaptive Quantizer', Eqs. (7)-(8); based on Clauset et al. power-law MLE and FQ-ViT log2 quantizer.
ad hoc to paper Calibration images sampled from the partially quantized model are representative of inference-time inputs.
Algorithm 1 lines 5-6; this is the core premise of PAR and is an unproven heuristic.

how reviews work

0 comments

Cite this review

Pith. "Pith review of TCAQ-DM: Timestep-Channel Adaptive Quantization for Diffusion Models." pith.science (2026). https://pith.science/paper/HJY7O7SW

@misc{pith2026241216700,
  author       = {Pith},
  title        = {Pith review of: TCAQ-DM: Timestep-Channel Adaptive Quantization for Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HJY7O7SW}},
  note         = {Machine review of arXiv:2412.16700}
}

read the original abstract

Diffusion models have achieved remarkable success in the image and video generation tasks. Nevertheless, they often require a large amount of memory and time overhead during inference, due to the complex network architecture and considerable number of timesteps for iterative diffusion. Recently, the post-training quantization (PTQ) technique has proved a promising way to reduce the inference cost by quantizing the float-point operations to low-bit ones. However, most of them fail to tackle with the large variations in the distribution of activations across distinct channels and timesteps, as well as the inconsistent of input between quantization and inference on diffusion models, thus leaving much room for improvement. To address the above issues, we propose a novel method dubbed Timestep-Channel Adaptive Quantization for Diffusion Models (TCAQ-DM). Specifically, we develop a timestep-channel joint reparameterization (TCR) module to balance the activation range along both the timesteps and channels, facilitating the successive reconstruction procedure. Subsequently, we employ a dynamically adaptive quantization (DAQ) module that mitigate the quantization error by selecting an optimal quantizer for each post-Softmax layers according to their specific types of distributions. Moreover, we present a progressively aligned reconstruction (PAR) strategy to mitigate the bias caused by the input mismatch. Extensive experiments on various benchmarks and distinct diffusion models demonstrate that the proposed method substantially outperforms the state-of-the-art approaches in most cases, especially yielding comparable FID metrics to the full precision model on CIFAR-10 in the W6A6 setting, while enabling generating available images in the W4A4 settings.

Figures

Figures reproduced from arXiv: 2412.16700 by the authors.

**Figure 1.** (a) Fluctuated activations per channels and timesteps in the convolutional layers (e.g. up.0.block.0.conv1 of DDIM). (b) Dynamic changes of activation distributions in the post-Softmax layer (e.g. down.1.attn.0 of DDIM) in distinct timesteps. (c) Misalignment between the intermediate data of the reconstruction stage in the quantization process and those in the inference process. Gao et al. 2023; Wang et al. 2024b)… view at source ↗

**Figure 2.** Overview of our proposed method. In the initialization stage, we develop the timestep-channel joint repa [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Visualization of images generated by quan [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 43 canonical work pages

[1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Castells, T.; Song, H.-K.; Kim, B.-K.; and Choi, S. 2024. LD-Pruner: Efficient pruning of latent diffusion models using task-agnostic insights. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 821--830

work page 2024
[4]

Chen, S.; Xu, M.; Ren, J.; Cong, Y.; He, S.; Xie, Y.; Sinha, A.; Luo, P.; Xiang, T.; and Perez-Rua, J.-M. 2024. GenTron: Diffusion transformers for image and video generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 6441--6451

work page 2024
[5]

Chu, X.; Li, L.; and Zhang, B. 2024. Make repvgg greater again: A quantization-qware approach. In Proceedings of the AAAI Conference on Artificial Intelligence, 10, 11624--11632

work page 2024
[6]

Chung, H.; Sim, B.; and Ye, J. C. 2022. Come-closer-diffuse-faster: Accelerating conditional diffusion models for inverse problems through stochastic contraction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 12413--12422

2022
[7]

R.; and Newman, M

Clauset, A.; Shalizi, C. R.; and Newman, M. E. 2009. Power-law distributions in empirical data. SIAM review, 661--703

work page 2009
[8]

Deng, J.; Dong, W.; Socher, R.; Li, L.; Li, K.; and Fei - Fei, L. 2009. ImageNet: A large-scale hierarchical image database. In Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition, 248--255

work page 2009

Show all 52 references

[9]

Dettmers, T.; Pagnoni, A.; Holtzman, A.; and Zettlemoyer, L. 2024. Qlora: Efficient finetuning of quantized llms. In Advances in Neural Information Processing Systems, 10088--10115

2024
[10]

Franzese, G.; Rossi, S.; Yang, L.; Finamore, A.; Rossi, D.; Filippone, M.; and Michiardi, P. 2023. How much is enough? A study on diffusion times in score-based generative models. Entropy, 633--643

2023
[11]

Gao, S.; Liu, X.; Zeng, B.; Xu, S.; Li, Y.; Luo, X.; Liu, J.; Zhen, X.; and Zhang, B. 2023. Implicit diffusion models for continuous super-resolution. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10021--10030

2023
[12]

Gong, R.; Liu, X.; Jiang, S.; Li, T.; Hu, P.; Lin, J.; Yu, F.; and Yan, J. 2019. Differentiable soft quantization: Bridging full-precision and low-bit neural networks. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 4852--4861

2019
[13]

He, Y.; Liu, L.; Liu, J.; Wu, W.; Zhou, H.; and Zhuang, B. 2024. Ptqd: Accurate post-training quantization for diffusion models. In Advances in Neural Information Processing Systems, 13237--13249

2024
[14]

Heusel, M.; Ramsauer, H.; Unterthiner, T.; Nessler, B.; and Hochreiter, S. 2017. Gans trained by a two time-scale update rule converge to a local nash equilibrium. In Advances in Neural Information Processing Systems, 6626--6637

2017
[15]

Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion probabilistic models. In Advances in Neural Information Processing Systems, 6840--6851

2020
[16]

Ho, J.; Salimans, T.; Gritsenko, A.; Chan, W.; Norouzi, M.; and Fleet, D. J. 2022. Video diffusion models. In Advances in Neural Information Processing Systems, 8633--8646

2022
[17]

Huang, T.; Zhang, Y.; Zheng, M.; You, S.; Wang, F.; Qian, C.; and Xu, C. 2024 a . Knowledge diffusion for distillation. In Advances in Neural Information Processing Systems, 65299--65316

2024
[18]

Huang, Y.; Gong, R.; Liu, J.; Chen, T.; and Liu, X. 2024 b . TFMQ-DM: Temporal feature maintenance quantization for diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 7362--7371

2024
[19]

Kim, B.; and Ye, J. C. 2023. Denoising MCMC for accelerating diffusion-based generative models. In Proceedings of the International Conference on Machine Learning, 16955--16977

2023
[20]

Krizhevsky, A.; Hinton, G.; et al. 2009. Learning multiple layers of features from tiny images. Technical report, Toronto, ON, Canada

2009
[21]

Kuzmin, A.; Van Baalen, M.; Ren, Y.; Nagel, M.; Peters, J.; and Blankevoort, T. 2022. Fp8 quantization: The power of the exponent. In Advances in Neural Information Processing Systems, 14651--14662

2022
[22]

Lam, M. W. Y.; Wang, J.; Su, D.; and Yu, D. 2022. BDDM: Bilateral denoising diffusion models for fast and high-quality speech synthesis. In Proceedings of the International Conference on Learning Representations

2022
[23]

Li, H.; Yang, Y.; Chang, M.; Chen, S.; Feng, H.; Xu, Z.; Li, Q.; and Chen, Y. 2022. Srdiff: Single image super-resolution with diffusion probabilistic models. Neurocomputing, 47--59

2022
[24]

Li, L.; Li, H.; Zheng, X.; Wu, J.; Xiao, X.; Wang, R.; Zheng, M.; Pan, X.; Chao, F.; and Ji, R. 2023 a . AutoDiffusion: Training-free optimization of time steps and architectures for automated diffusion model acceleration. In Proceedings of the IEEE/CVF International Conferenc...

2023
[25]

Li, X.; Liu, Y.; Lian, L.; Yang, H.; Dong, Z.; Kang, D.; Zhang, S.; and Keutzer, K. 2023 b . Q-diffusion: Quantizing diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 17535--17545

2023
[26]

Li, Y.; Gong, R.; Tan, X.; Yang, Y.; Hu, P.; Zhang, Q.; Yu, F.; Wang, W.; and Gu, S. 2021. BRECQ: Pushing the limit of post-training quantization by block reconstruction. In Proceedings of the International Conference on Learning Representations

2021
[27]

Li, Z.; Xiao, J.; Yang, L.; and Gu, Q. 2023 c . Repq-vit: Scale reparameterization for post-training quantization of vision transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 17227--17236

2023
[28]

Lin, Y.; Zhang, T.; Sun, P.; Li, Z.; and Zhou, S. 2022. FQ-ViT: Post-training quantization for fully quantized vision transformer. In Proceedings of the International Joint Conference on Artificial Intelligence, 1173--1179

2022
[29]

Luhman, E.; and Luhman, T. 2021. Knowledge distillation in iterative generative models for improved sampling speed. arXiv preprint arXiv:2101.02388

2021 arXiv
[30]

Lyu, Z.; Xu, X.; Yang, C.; Lin, D.; and Dai, B. 2022. Accelerating diffusion models via early stop of the diffusion process. arXiv preprint arXiv:2205.12524

2022 arXiv
[31]

Ma, X.; Fang, G.; and Wang, X. 2024. Deepcache: Accelerating diffusion models for free. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 15762--15772

2024
[32]

A.; Van Baalen, M.; Louizos, C.; and Blankevoort, T

Nagel, M.; Amjad, R. A.; Van Baalen, M.; Louizos, C.; and Blankevoort, T. 2020. Up or down? Adaptive rounding for post-training quantization. In Proceedings of the International Conference on Machine Learning, 7197--7206

2020
[33]

Salimans, T.; Goodfellow, I.; Zaremba, W.; Cheung, V.; Radford, A.; and Chen, X. 2016. Improved techniques for training gans. In Advances in Neural Information Processing Systems, 2226--2234

2016
[34]

Salimans, T.; and Ho, J. 2022. Progressive distillation for fast sampling of diffusion models. In Proceedings of the International Conference on Learning Representations

2022
[35]

Shang, Y.; Yuan, Z.; Xie, B.; Wu, B.; and Yan, Y. 2023. Post-training quantization on diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 1972--1981

2023
[36]

Song, J.; Meng, C.; and Ermon, S. 2021. Denoising diffusion implicit models. In Proceedings of the International Conference on Learning Representations

2021
[37]

Su, X.; Song, J.; Meng, C.; and Ermon, S. 2023. Dual diffusion implicit bridges for image-to-image translation. In Proceedings of the International Conference on Learning Representations

2023
[38]

Sun, H.; Tang, C.; Wang, Z.; Meng, Y.; Jiang, J.; Ma, X.; and Zhu, W. 2024. TMPQ-DM: Joint timestep reduction and quantization precision selection for efficient diffusion models. arXiv preprint arXiv:2404.09532

2024 arXiv
[39]

Tang, S.; Wang, X.; Chen, H.; Guan, C.; Wu, Z.; Tang, Y.; and Zhu, W. 2024. Post-training quantization with progressive calibration and activation relaxing for text-to-image diffusion models. In Proceedings of the European Conference on Computer Vision, 404--420

2024
[40]

Tumanyan, N.; Geyer, M.; Bagon, S.; and Dekel, T. 2023. Plug-and-play diffusion features for text-driven image-to-image translation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 1921--1930

2023
[41]

Wang, C.; Wang, Z.; Xu, X.; Tang, Y.; Zhou, J.; and Lu, J. 2024 a . Towards accurate post-training quantization for diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 16026--16035

2024
[42]

C.; and Wen, B

Wang, Y.; Yang, W.; Chen, X.; Wang, Y.; Guo, L.; Chau, L.-P.; Liu, Z.; Qiao, Y.; Kot, A. C.; and Wen, B. 2024 b . SinSR: Diffusion-based image super-resolution in a single step. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 25796--25805

2024
[43]

Watson, D.; Chan, W.; Ho, J.; and Norouzi, M. 2022. Learning fast samplers for diffusion models by differentiating through sample quality. In Proceedings of the International Conference on Learning Representations

2022
[44]

Wei, X.; Gong, R.; Li, Y.; Liu, X.; and Yu, F. 2022. QDrop: Randomly dropping quantization for extremely low-bit post-training quantization. In Proceedings of the International Conference on Learning Representations

2022
[45]

Wu, Z.; Chen, J.; Zhong, H.; Huang, D.; and Wang, Y. 2024. AdaLog: Post-training quantization for vision transformers with adaptive logarithm quantizer. In Proceedings of the European Conference on Computer Vision, 411--427

2024
[46]

Yao, Y.; Tian, F.; Chen, J.; Lin, H.; Dai, G.; Liu, Y.; and Wang, J. 2024. Timestep-aware correction for quantized diffusion models. In Proceedings of the European Conference on Computer Vision, 215--232

2024
[47]

Yu, F.; Zhang, Y.; Song, S.; Seff, A.; and Xiao, J. 2015. LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop. arXiv preprint arXiv:1506.03365

2015 arXiv
[48]

Zhang, D.; Li, S.; Chen, C.; Xie, Q.; and Lu, H. 2024. Laptop-diff: Layer pruning and normalized distillation for compressing diffusion models. arXiv preprint arXiv:2404.11098

2024 arXiv
[49]

Zhang, L.; He, Y.; Lou, Z.; Ye, X.; Wang, Y.; and Zhou, H. 2023. Root quantization: A self-adaptive supplement STE . Appl. Intell., 6266--6275

2023
[50]

Zhang, Q.; and Chen, Y. 2023. Fast sampling of diffusion models with exponential integrator. In Proceedings of the International Conference on Learning Representations

2023
[51]

Zhao, W.; Bai, L.; Rao, Y.; Zhou, J.; and Lu, J. 2024. Unipc: A unified predictor-corrector framework for fast sampling of diffusion models. In Advances in Neural Information Processing Systems, 49842--49869

2024
[52]

Zhou, Z.; Chen, D.; Wang, C.; and Chen, C. 2024. Fast ode-based sampling for diffusion models in around 5 steps. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 7777--7786

2024

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Castells, T.; Song, H.-K.; Kim, B.-K.; and Choi, S. 2024. LD-Pruner: Efficient pruning of latent diffusion models using task-agnostic insights. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 821--830

work page 2024

[4] [4]

Chen, S.; Xu, M.; Ren, J.; Cong, Y.; He, S.; Xie, Y.; Sinha, A.; Luo, P.; Xiang, T.; and Perez-Rua, J.-M. 2024. GenTron: Diffusion transformers for image and video generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 6441--6451

work page 2024

[5] [5]

Chu, X.; Li, L.; and Zhang, B. 2024. Make repvgg greater again: A quantization-qware approach. In Proceedings of the AAAI Conference on Artificial Intelligence, 10, 11624--11632

work page 2024

[6] [6]

Chung, H.; Sim, B.; and Ye, J. C. 2022. Come-closer-diffuse-faster: Accelerating conditional diffusion models for inverse problems through stochastic contraction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 12413--12422

2022

[7] [7]

R.; and Newman, M

Clauset, A.; Shalizi, C. R.; and Newman, M. E. 2009. Power-law distributions in empirical data. SIAM review, 661--703

work page 2009

[8] [8]

Deng, J.; Dong, W.; Socher, R.; Li, L.; Li, K.; and Fei - Fei, L. 2009. ImageNet: A large-scale hierarchical image database. In Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition, 248--255

work page 2009