mlsanityData to 14 Oct 2026

MPQ-Diff: Mixed Precision Quantization for Diffusion Models

arXiv 2412.00144 ↗First posted 28 Nov 2024cs.CV cs.LGcited by 1

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

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.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 15 of its 29 references that are in the map. First posted on arXiv 28 Nov 2024.

Builds on

15 of 29 references.

Closest ancestors · 5

Same direction, newest first

  1. 2023Temporal Dynamic Quantization for Diffusion Models4 Jun
  2. 2023PTQD: Accurate Post-Training Quantization for Diffusion Models18 May
  3. 2022Post-training Quantization on Diffusion Models28 Nov
  4. 2022Understanding Diffusion Models: A Unified Perspective25 Aug
  5. 2020Improved Techniques for Training Score-Based Generative Models16 Jun

Other references · 10

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  2. 2019Similarity of Neural Network Representations Revisitedcited by 1,755
  3. 2015LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loopcited by 1,355
  4. 2020Up or Down? Adaptive Rounding for Post-Training Quantizationcited by 621
  5. 2018HAQ: Hardware-Aware Automated Quantization with Mixed Precisioncited by 443
  6. 2019HAWQ-V2: Hessian Aware trace-Weighted Quantization of Neural Networkscited by 261
  7. 2019Differentiable Soft Quantization: Bridging Full-Precision and Low-Bit Neural Networkscited by 238
  8. 2022QDrop: Randomly Dropping Quantization for Extremely Low-bit Post-Training Quantizationcited by 180
  9. 2020Improving Post Training Neural Quantization: Layer-wise Calibration and Integer Programmingcited by 92
  10. 2020FracBits: Mixed Precision Quantization via Fractional Bit-Widthscited by 55

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026TASQ: Temporal-Adaptive Bit Sparsification Quantization for Diffusion Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024MPQ-DM: Mixed Precision Quantization for Extremely Low Bit Diffusion Models
  2. 2026TMPDiff: Temporal Mixed-Precision for Diffusion Models
  3. 2025DMQ: Dissecting Outliers of Diffusion Models for Post-Training Quantization
  4. 2025MPQ-DMv2: Flexible Residual Mixed Precision Quantization for Low-Bit Diffusion Models with Temporal Distillation
  5. 2024PQD: Post-training Quantization for Efficient Diffusion Models
  6. 2024Qua2SeDiMo: Quantifiable Quantization Sensitivity of Diffusion Models
  7. 2025AccuQuant: Simulating Multiple Denoising Steps for Quantizing Diffusion Models
  8. 2024TMPQ-DM: Joint Timestep Reduction and Quantization Precision Selection for Efficient Diffusion Models