Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

MPQ-Diff: Mixed Precision Quantization for Diffusion Models

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read MPQ-Diff: per-layer bit allocation beats uniform quantization for diffusion models

desk verdict Timestep-aware ORM bit allocation for diffusion models: same-budget FID wins in all six Table 1 rows, but the headline numbers are a different-size comparison and the ORM-sensitivity link is never tested; worth a serious referee. read the letter →

arxiv 2412.00144 v1 pith:BUJ6OVNR submitted 2024-11-28 cs.CV cs.LG

Rocco Manz Maruzzelli , Basile Lewandowski , Lydia Y. Chen This is my paper · ORCID
classification cs.CVcs.LG
keywords mixedprecisionquantizationdiffusionmodelspost-trainingnetworkorthogonalityFIDtimestepaggregationlinearprogrammingU-Net
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 are slow because every denoising step runs a large U-Net, and low-bit quantization speeds them up but degrades image quality. MPQ-Diff asks whether giving different layers different bit-widths—rather than one uniform width—can recover that quality at the same memory budget. The paper's answer is yes. It treats a layer's average cross-correlation with all other layers, called the network orthogonality metric (equivalent to linear CKA), as a proxy for how much that layer matters, aggregates this across sampled denoising timesteps, and solves a small linear program to allocate 2–8 bit weights per layer. On ImageNet and LSUN benchmarks, the resulting mixed-precision models improve FID by 1.6–2.85 points at the same model size, and by about 50 points when a small extra memory allowance is granted.

What carries the argument

The load-bearing object is the ORthogonality Metric (ORM), the Frobenius inner product of two layers' activations normalized by their norms, which equals the linear Centered Kernel Alignment (CKA) similarity. The paper aggregates ORM over time by zero-normalizing each layer's row sum $\gamma_i^{(t)}$ across timesteps, exponentiating the negative z-score to form weights $\theta_i$, and feeding these into a linear program whose objective (Equation 7) assigns higher bit-widths to more independent layers under a model-size constraint (Equation 8), with uniform sampling of timesteps (Section 4.3) to control profiling cost.

What would settle it

Compare MPQ-Diff's bit allocations against allocations chosen by directly measured per-layer output error (e.g., reconstruction MSE or Hessian trace) on the same LDM-4 model; if ORM-ranked layers do not match error-ranked layers, the FID gains would be shown to come from the LPP flexibility rather than the metric. A simpler control: shuffle the ORM scores before solving the LPP; if FID stays similar, ORM is not the driver.

Watch

Extended reading notes

Core claim

The paper claims that a per-layer mixed-precision bit allocation, chosen by maximizing a time-aggregated orthogonality objective under a memory constraint, consistently outperforms uniform 4-bit quantization of diffusion models at the same or slightly larger model size. The central quantity is the ORM of Equation 5, which gives a value in $[0,1]$ measuring how much two layers' outputs correlate; a layer's weight is its summed ORM against all other layers, normalized and exponentially transformed so that high-importance (low-correlation) layers receive more bits. The allocation is obtained by solving a linear programming problem, and to keep profiling costs low the ORM is computed on a uniformly sampled subset of denoising steps rather than all of them.

Load-bearing premise

The allocation is only as good as the orthogonality metric's ability to rank layers by how much quantization hurts them; the paper adopts ORM from classification-network work and does not validate that it tracks actual per-layer quantization error in U-Nets with skip connections.

Editorial extensions

If this is right

  • Because MPQ-Diff wraps any fixed-precision post-training quantization method, both PTQD and EfficientDM can inherit the FID improvement without additional training.
  • The bit-allocation procedure is one-shot: a single forward pass over a batch gives the ORM matrices, then the linear program runs in negligible time, so different memory budgets can be explored cheaply.
  • The FID-versus-model-size curve becomes a smooth, tunable knob rather than a few discrete uniform-width points, as shown by the ablation in Figure 7.
  • With only a ~16% increase in model size over 2-bit uniform quantization, mixed precision reaches FID 15.39 on LSUN-Churches, while 4-bit uniform needs 29% more memory to reach FID 9.75.
  • Sampling one quarter of the denoising timesteps keeps the allocation within about 3% of using all timesteps, so the method scales to models with hundreds of sampling steps.

Reading between the lines

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

  • Because ORM equals linear CKA, the method implicitly assumes representational similarity to other layers tracks quantizability; this assumption is untested for U-Nets with skip connections, so a direct comparison against per-layer reconstruction-error or Hessian-based sensitivity scores would be the natural next experiment.
  • The same time-aggregation and linear-programming recipe could transfer to other iterative generative models that reuse a single network across steps, such as flow matching or consistency models.
  • The stability of the allocation under uniform sampling suggests that a timestep-dependent allocation, where bit-widths change mid-sampling, might yield further quality gains at the same memory cost.
  • If ORM rankings were shuffled before solving the LPP and FID stayed similar, that would indicate the gains come from the optimization flexibility rather than the orthogonality proxy itself; this is a testable control experiment.
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.

Referee Report

4 major / 5 minor

Summary. The paper proposes MPQ-Diff, a post-training mixed-precision quantization scheme for diffusion models. It computes a per-layer 'network orthogonality metric' (ORM), defined in Eq. (5) and equivalent to linear CKA, at sampled timesteps; aggregates these scores using an exponential weighting; and feeds them into a linear programming problem inherited from OMPQ to select bit widths per layer under a model-size budget. Experiments on ImageNet, LSUN-Churches, and LSUN-Bedrooms with LDM-4/LDM-8 show same-size FID improvements over PTQD and EfficientDM in most settings, and larger-model-size versions show large FID drops from a fixed 2-bit baseline. The paper also proposes uniform timestep sampling to reduce profiling cost and reports an ablation of sampling fractions.

Significance. If the central assumption were validated, the work would be a useful first mixed-precision post-training quantization scheme for diffusion models, and the same-size FID gains in Table 1 are plausible evidence that non-uniform bit allocation can help. The integration with two existing fixed-precision methods (PTQD and EfficientDM) is a strength. However, the paper does not release code or trained bit allocations, and its significance hinges on the untested claim that ORM ranks per-layer sensitivity to quantization error in diffusion U-Nets; as submitted, the evidence is insufficient to establish that, and the headline numbers are not equal-size comparisons. The paper is a reasonable starting point, but the core mechanism needs direct validation before the contribution can be assessed.

major comments (4)
  1. [§3.3, Eq. (5); Algorithm 1] The central claim rests on the assumption that ORM, defined in Eq. (5) and equal to linear CKA, ranks the sensitivity of layers to quantization error. This assumption is inherited from OMPQ, which was designed for classification networks, and is not validated for diffusion U-Nets with skip connections. The paper provides no experiment comparing the ORM ranking with per-layer quantization error, nor an ablation replacing the ORM scores with random or reversed importance scores. Such a sensitivity-rank check or randomized baseline is necessary to establish that the bit allocation, not just the added flexibility of mixed precision, produces the observed FID gains.
  2. [Abstract; Introduction (p.1); §5.3, Figure 7] The headline improvements (FID 65.73→15.39 and 52.66→14.93) are not same-size comparisons: they compare a fixed 2-bit model at 109 MB with a mixed-precision model at 127 MB (16.5% larger), as correctly stated in Section 5.3. The abstract reports these numbers 'compared to their fixed precision quantization' without mentioning the size difference, and the Introduction says 'only 10% increase in model size,' contradicting the 16.5% figure. The equal-size comparisons in Table 1 are the relevant evidence; the Figure 7 comparisons should be relabeled or removed from the abstract.
  3. [Table 1; §5.1] On ImageNet, Ours(EfficientDM) improves FID from 8.64 to 6.07 but sFID degrades sharply from 7.82 to 11.46. The text in Section 5.1 says sFID does 'not show a similar performance' but does not quantify this regression. Given the paper's claim of 'consistent improvements' over fixed-precision quantization, the sFID degradation on a main benchmark is a substantive caveat that should be analyzed rather than dismissed.
  4. [Introduction (p.1); §3.3; Algorithm 1] The text is internally inconsistent about the sign of the importance score. The Introduction states that a layer with higher aggregated ORM has higher relative importance, while Section 3.3 says larger bit-widths should go to layers with stronger orthogonality (i.e., lower ORM), and Algorithm 1's θ_i = e^{-ρ_i} implements the latter. This contradiction needs to be resolved: either the prose or the algorithm is wrong, and the paper should state explicitly that low ORM corresponds to high importance.
minor comments (5)
  1. [§3.2] The quantization formula uses ⌊·⌋ but calls it 'the round operation'; use round() or define the rounding explicitly.
  2. [Figure 7] The legend 'Ours Uniform quantization 2-bit 4-bit' is unclear; distinguish points for fixed 2-bit, fixed 4-bit, and mixed precision with model sizes labeled.
  3. [§5.3] The text says 'only a slight increase in model size' and elsewhere '10% increase,' while the actual 16.5% figure appears only in one place; the inconsistency should be fixed throughout.
  4. [§3.3] The term 'network orthogonality metric' is potentially confusing because it is a correlation measure; consider renaming to 'cross-layer correlation' or defining the direction of orthogonality clearly.
  5. [§5] The paper does not report wall-clock time or memory overhead of the profiling step (including the 50-timestep sampling), which is relevant for the claimed efficiency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the bit allocation is computed from activation-based ORM/CKA scores and a borrowed LPP, then evaluated on held-out FID; no fitted parameter is renamed as a prediction.

full rationale

I walked the derivation chain from Eq. (5) through Algorithm 1 and Eqs. (7)-(8). The ORM score is a known metric (the paper explicitly notes it is the same as linear CKA), and the per-layer importance theta is computed from activation statistics alone, independent of FID or sFID. The LPP objective and constraints are inherited from OMPQ [19], which is a cited external work by different authors, not a self-citation; no load-bearing premise is justified by the present authors' own prior results. The FID improvements in Tables 1 and Figure 7 are measured after the allocation is fixed, so the headline gains are empirical outcomes rather than consequences of the definition of theta. The reader's concerns about the unvalidated transfer of the orthogonality-to-quantization-sensitivity assumption from classification networks, the internal sign inconsistency between Section 3.3 and Algorithm 1, and the mismatched model sizes in Figure 7 are substantive correctness or reporting risks, but they are not circularity: the paper does not fit a parameter to FID and then report that fit as a prediction, nor does any equation reduce to its own input by construction. Therefore no circular step can be exhibited, and the honest finding is score 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central allocation rests on heuristics borrowed from OMPQ: the ORM importance proxy, the LPP objective, and several hand-chosen aggregation functions. No new physical or architectural entities are introduced.

free parameters (5)
  • Time-aggregation weighting w_i = exp(-ZScore(gamma_i)) = not reported
    Hand-chosen exponential weighting with no justification; affects which timesteps dominate the allocation.
  • Exponential mapping theta_i = exp(-rho_i) = not reported
    Hand-chosen nonlinearity; determines conversion from aggregated ORM to LPP importance weights.
  • Model size target T in LPP constraint = e.g., 250, 194, 181, 127 MB
    The target size is chosen per experiment, often to match fixed 4-bit models; the 'dramatic' FID improvements come from a different T (127 MB vs 109 MB).
  • Bit-width range B = 1-8 bits (footnote '18 bits' is ambiguous)
    Allowed bit-width set for weights; the paper does not specify the exact discrete set used.
  • Timestep sampling fraction = 1/4 (50 of 100/200 steps)
    The 1/4 fraction is selected based on Table 2 on the same evaluation datasets, making it a data-dependent choice.
assumptions (4)
  • domain assumption ORM (linear CKA) is a valid proxy for layer importance in quantization
    Borrowed from OMPQ [19] for classification; not re-validated for diffusion U-Nets.
  • domain assumption The LPP objective from OMPQ (Equation 7) maximizes representation capability of the quantized model
    The paper uses OMPQ's objective without deriving why this allocation rule is optimal.
  • domain assumption Activations of neighboring timesteps are similar enough that uniform sampling preserves layer importance
    Supported only by Figure 5 for LDM-4 on ImageNet; no theoretical guarantee.
  • domain assumption Quantizing to lower bit-width while maintaining FID is the correct optimization target
    The paper assumes FID is the right metric, though sFID results sometimes worsen.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MPQ-Diff: Mixed Precision Quantization for Diffusion Models." pith.science (2026). https://pith.science/paper/BUJ6OVNR

@misc{pith2026241200144,
  author       = {Pith},
  title        = {Pith review of: MPQ-Diff: Mixed Precision Quantization for Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BUJ6OVNR}},
  note         = {Machine review of arXiv:2412.00144}
}
read the original abstract

Diffusion models (DMs) generate remarkable high quality images via the stochastic denoising process, which unfortunately incurs high sampling time. Post-quantizing the trained diffusion models in fixed bit-widths, e.g., 4 bits on weights and 8 bits on activation, is shown effective in accelerating sampling time while maintaining the image quality. Motivated by the observation that the cross-layer dependency of DMs vary across layers and sampling steps, we propose a mixed precision quantization scheme, MPQ-Diff, which allocates different bit-width to the weights and activation of the layers. We advocate to use the cross-layer correlation of a given layer, termed network orthogonality metric, as a proxy to measure the relative importance of a layer per sampling step. We further adopt a uniform sampling scheme to avoid the excessive profiling overhead of estimating orthogonality across all time steps. We evaluate the proposed mixed-precision on LSUN and ImageNet, showing a significant improvement in FID from 65.73 to 15.39, and 52.66 to 14.93, compared to their fixed precision quantization, respectively.

Figures

Figures reproduced from arXiv: 2412.00144 by the authors.

Figure 1
Figure 1. Samples of generated images under different bit precision. other layers, indicating the higher relative importance com￾pared to the layers with lower aggregated ORM. We adapt this concept to the iterative structure of DMs, developing a timestep-aware bit-width allocation method that accounts for the varying importance of layers, approximated by their aggregated ORM, across the denoising steps. These metrics are then… view at source ↗
Figure 2
Figure 2. Overview of the MPQ-Diff workflow. a) Deconstruct the DM into a set of functions F , which are used across all T generation timesteps. b) The ORM matrices for every sampled timestep is calculated from F . c) Aggregation of all ORM matrices to obtain overall function importance across timesteps. d) LPP constructed by the importance factor θ to derive bit configuration. Timestep 1 Timestep 11 Timestep 20 0.0 0.2 0.4 0… view at source ↗
Figure 3
Figure 3. Orthogonality Matrices across timesteps, for LDM-4 on [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (18 more)
Figure 4
Figure 4. Figure 4: Activation ranges of xt across all 100 time steps of LDM-4 model on Imagenet 256 × 256. The blue regions represent the inter-quartile range (first to third quartile) of activation values, while the gray regions extend from the 5th to 95th percentiles. as was outlined i…
Figure 6
Figure 6. Figure 6: Different γ coefficients (one for each timestep) per layer on LDM-4 Imagenet (steps = 20 eta = 0.0 scale = 3.0). It appears the inner layers show smaller variance over time. els with more than 50 timesteps, we implement a uniform sampling method to select a subset of t…
Figure 7
Figure 7. Figure 7: FID for model sizes on LDM-4 LSUN-Churches, [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Quantization Learning Process of LSUN-Churches LDM-8 [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: UNIFORM 4 bits – 181 Mb, LSUN-Churches 256 [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: Mixed Precision – 181 Mb, LSUN-Churches 256 3 [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: UNIFORM 4 bits – 194 Mb, LSUN-Bedrooms LDM-4 (steps [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]
Figure 12
Figure 12. Figure 12: Mixed Precision – 194 Mb, LSUN-Bedrooms LDM-4 (steps 4 [PITH_FULL_IMAGE:figures/full_fig_p014_12.png]
Figure 13
Figure 13. Figure 13: UNIFORM 4 bits – 250 Mb, ImageNet LDM-4 (steps [PITH_FULL_IMAGE:figures/full_fig_p015_13.png]
Figure 14
Figure 14. Figure 14: Mixed Precision – 250 Mb, ImageNet LDM-4 (steps 5 [PITH_FULL_IMAGE:figures/full_fig_p015_14.png]
Figure 15
Figure 15. Figure 15: UNIFORM 4 bits – 181 Mb, LSUN-Churches 256 [PITH_FULL_IMAGE:figures/full_fig_p016_15.png]
Figure 16
Figure 16. Figure 16: Mixed Precision – 181 Mb, LSUN-Churches 256 6 [PITH_FULL_IMAGE:figures/full_fig_p016_16.png]
Figure 17
Figure 17. Figure 17: UNIFORM 4 bits – 194 Mb, LSUN-Bedrooms LDM-4 (steps [PITH_FULL_IMAGE:figures/full_fig_p017_17.png]
Figure 18
Figure 18. Figure 18: Mixed Precision – 194 Mb, LSUN-Bedrooms LDM-4 (steps 7 [PITH_FULL_IMAGE:figures/full_fig_p017_18.png]
Figure 19
Figure 19. Figure 19: UNIFORM 4 bits – 250 Mb, ImageNet LDM-4 (steps [PITH_FULL_IMAGE:figures/full_fig_p018_19.png]
Figure 20
Figure 20. Figure 20: Mixed Precision – 250 Mb, ImageNet LDM-4 (steps 8 [PITH_FULL_IMAGE:figures/full_fig_p018_20.png]
Figure 21
Figure 21. Figure 21: Comparison of images using differing sized models as in [PITH_FULL_IMAGE:figures/full_fig_p019_21.png]
Figure 22
Figure 22. Figure 22: Mean square error between quantized and full precision image generation [PITH_FULL_IMAGE:figures/full_fig_p020_22.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. TASQ: Temporal-Adaptive Bit Sparsification Quantization for Diffusion Models

    cs.CV 2026-08 conditional novelty 7.0 of 10

    A temporal-spatial LSB mask over one shared weight buffer lets diffusion models use lower bit precision in less sensitive denoising stages, cutting compute by 25-50% on bit-serial hardware with no loss in image quality.

Reference graph

Works this paper leans on

38 extracted references · 35 canonical work pages · cited by 1 Pith paper

  1. [1]

    Aca- demic press, 2011

    George B Arfken, Hans J Weber, and Frank E Harris.Mathe- matical methods for physicists: a comprehensive guide. Aca- demic press, 2011. 3

    work page 2011

  2. [2]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE Conference on Computer Vision and Pattern Recognition, pages 248–255, 2009. 7

    2009

  3. [3]

    Di ffusion models beat gans on image synthesis, 2021

    Prafulla Dhariwal and Alex Nichol. Di ffusion models beat gans on image synthesis, 2021. 1, 7

    work page 2021

  4. [4]

    Mahoney, and Kurt Keutzer

    Zhen Dong, Zhewei Yao, Yaohui Cai, Daiyaan Arfeen, Amir Gholami, Michael W. Mahoney, and Kurt Keutzer. Hawq- v2: Hessian aware trace-weighted quantization of neural net- works, 2019. 1

    work page 2019

  5. [5]

    Esser, Je ffrey L

    Steven K. Esser, Je ffrey L. McKinstry, Deepika Bablani, Rathinakumar Appuswamy, and Dharmendra S. Modha. Learned step size quantization, 2020. 2

    work page 2020

  6. [6]

    Structural pruning for diffusion models, 2023

    Gongfan Fang, Xinyin Ma, and Xinchao Wang. Structural pruning for diffusion models, 2023. 1

    work page 2023

  7. [7]

    Differen- tiable soft quantization: Bridging full-precision and low-bit neural networks, 2019

    Ruihao Gong, Xianglong Liu, Shenghu Jiang, Tianxiang Li, Peng Hu, Jiazhen Lin, Fengwei Yu, and Junjie Yan. Differen- tiable soft quantization: Bridging full-precision and low-bit neural networks, 2019. 2

    work page 2019

  8. [8]

    Ptqd: Accurate post-training quantization for diffusion models, 2023

    Yefei He, Luping Liu, Jing Liu, Weijia Wu, Hong Zhou, and Bohan Zhuang. Ptqd: Accurate post-training quantization for diffusion models, 2023. 1, 2, 4, 6, 7

    work page 2023

Show all 38 references
  1. [9]

    E fficientdm: E fficient quantization-aware fine- tuning of low-bit diffusion models, 2024

    Yefei He, Jing Liu, Weijia Wu, Hong Zhou, and Bohan Zhuang. E fficientdm: E fficient quantization-aware fine- tuning of low-bit diffusion models, 2024. 1, 2, 7

    2024

  2. [10]

    Gans trained by a two time-scale update rule converge to a local nash equilib- rium, 2018

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilib- rium, 2018. 7

    2018

  3. [11]

    Denoising di ffu- sion probabilistic models, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising di ffu- sion probabilistic models, 2020. 1, 2

    2020

  4. [12]

    Improving post training neural quantization: Layer-wise calibration and integer programming, 2020

    Itay Hubara, Yury Nahshan, Yair Hanani, Ron Banner, and Daniel Soudry. Improving post training neural quantization: Layer-wise calibration and integer programming, 2020. 2

    2020

  5. [13]

    Similarity of neural network representa- tions revisited

    Simon Kornblith, Mohammad Norouzi, Honglak Lee, and Geoffrey Hinton. Similarity of neural network representa- tions revisited. In Proceedings of the 36th International Conference on Machine Learning, pages 3519–3529. PMLR,

  6. [14]

    A software package for sequential quadratic programming

    Dieter Kraft. A software package for sequential quadratic programming. Technical Report DFVLR-FB 88-28, DLR German Aerospace Center — Institute for Flight Mechanics, Cologne, Germany, 1988. 6

    1988

  7. [15]

    Q-diffusion: Quantizing diffusion models, 2023

    Xiuyu Li, Yijiang Liu, Long Lian, Huanrui Yang, Zhen Dong, Daniel Kang, Shanghang Zhang, and Kurt Keutzer. Q-diffusion: Quantizing diffusion models, 2023. 1, 2, 6

    2023

  8. [16]

    Brecq: Pushing the limit of post-training quantization by block reconstruc- tion, 2021

    Yuhang Li, Ruihao Gong, Xu Tan, Yang Yang, Peng Hu, Qi Zhang, Fengwei Yu, Wei Wang, and Shi Gu. Brecq: Pushing the limit of post-training quantization by block reconstruc- tion, 2021. 1, 2, 4

    2021

  9. [17]

    Relaxed quantization for discretized neural networks, 2018

    Christos Louizos, Matthias Reisser, Tijmen Blankevoort, Ef- stratios Gavves, and Max Welling. Relaxed quantization for discretized neural networks, 2018. 2

    2018

  10. [18]

    Understanding di ffusion models: A unified per- spective, 2022

    Calvin Luo. Understanding di ffusion models: A unified per- spective, 2022. 1, 3

    2022

  11. [19]

    Ompq: Orthogonal mixed precision quantization, 2022

    Yuexiao Ma, Taisong Jin, Xiawu Zheng, Yan Wang, Huixia Li, Yongjian Wu, Guannan Jiang, Wei Zhang, and Rongrong Ji. Ompq: Orthogonal mixed precision quantization, 2022. 1, 2, 3, 4, 5

    2022

  12. [20]

    Up or down? adaptive rounding for post-training quantization, 2020

    Markus Nagel, Rana Ali Amjad, Mart van Baalen, Christos Louizos, and Tijmen Blankevoort. Up or down? adaptive rounding for post-training quantization, 2020. 2

    2020

  13. [21]

    Overcoming oscillations in quantization-aware training, 2022

    Markus Nagel, Marios Fournarakis, Yelysei Bondarenko, and Tijmen Blankevoort. Overcoming oscillations in quantization-aware training, 2022. 2

    2022

  14. [22]

    Improved denoising dif- fusion probabilistic models, 2021

    Alex Nichol and Prafulla Dhariwal. Improved denoising dif- fusion probabilistic models, 2021. 1, 4

    2021

  15. [23]

    High-resolution image syn- thesis with latent diffusion models, 2022

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. High-resolution image syn- thesis with latent diffusion models, 2022. 1, 7

    2022

  16. [24]

    U- net: Convolutional networks for biomedical image segmen- tation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- net: Convolutional networks for biomedical image segmen- tation. In Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015 , pages 234–241, Cham, 2015. Springer International Publishing. 3

    2015

  17. [25]

    Progressive distillation for fast sampling of diffusion models, 2022

    Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models, 2022. 1

    2022

  18. [26]

    Improved techniques for training gans, 2016

    Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, and Xi Chen. Improved techniques for training gans, 2016. 7

    2016

  19. [27]

    Post-training quantization on diffusion models

    Yuzhang Shang, Zhihang Yuan, Bin Xie, Bingzhe Wu, and Yan Yan. Post-training quantization on diffusion models. In CVPR, 2023. 1, 2

    2023

  20. [28]

    Temporal dynamic quantization for dif- fusion models, 2023

    Junhyuk So, Jungwon Lee, Daehyun Ahn, Hyungjun Kim, and Eunhyeok Park. Temporal dynamic quantization for dif- fusion models, 2023. 1

    2023

  21. [29]

    Denois- ing diffusion implicit models, 2022

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models, 2022. 1, 2

    2022

  22. [30]

    Improved techniques for training score-based generative models, 2020

    Yang Song and Stefano Ermon. Improved techniques for training score-based generative models, 2020. 1

    2020

  23. [31]

    Haq: Hardware-aware automated quantization with mixed precision

    Kuan Wang, Zhijian Liu, Yujun Lin, Ji Lin, and Song Han. Haq: Hardware-aware automated quantization with mixed precision. In Proceedings of the IEEE /CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 2

    2019

  24. [32]

    Haq: Hardware-aware automated quantization with mixed precision, 2019

    Kuan Wang, Zhijian Liu, Yujun Lin, Ji Lin, and Song Han. Haq: Hardware-aware automated quantization with mixed precision, 2019. 1

    2019

  25. [33]

    Patch di ffusion: Faster and more data-efficient training of diffusion models

    Zhendong Wang, Yifan Jiang, Huangjie Zheng, Peihao Wang, Pengcheng He, Zhangyang ”Atlas” Wang, Weizhu Chen, and Mingyuan Zhou. Patch di ffusion: Faster and more data-efficient training of diffusion models. In Advances in Neural Information Processing Systems , pages 72137– 721...

    2023

  26. [34]

    Di ffusion-gan: Training gans with diffusion, 2023

    Zhendong Wang, Huangjie Zheng, Pengcheng He, Weizhu Chen, and Mingyuan Zhou. Di ffusion-gan: Training gans with diffusion, 2023. 1

    2023

  27. [35]

    Qdrop: Randomly dropping quantization for extremely low-bit post-training quantization, 2023

    Xiuying Wei, Ruihao Gong, Yuhang Li, Xianglong Liu, and Fengwei Yu. Qdrop: Randomly dropping quantization for extremely low-bit post-training quantization, 2023. 2 9

    2023

  28. [36]

    Fracbits: Mixed precision quanti- zation via fractional bit-widths, 2020

    Linjie Yang and Qing Jin. Fracbits: Mixed precision quanti- zation via fractional bit-widths, 2020. 1

    2020

  29. [37]

    Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop, 2016

    Fisher Yu, Ari Se ff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop, 2016. 7

    2016

  30. [38]

    Towards e ffective low-bitwidth convolutional neural networks, 2017

    Bohan Zhuang, Chunhua Shen, Mingkui Tan, Lingqiao Liu, and Ian Reid. Towards e ffective low-bitwidth convolutional neural networks, 2017. 2 10 MPQ-Diff: Mixed Precision Quantization for Di ffusion Models Supplementary Material 1 Figure 8. Quantization Learning Process of LSUN-...

    2017

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.