OrbitQuant: Data-Agnostic Quantization for Image and Video Diffusion Transformers
Summary
OrbitQuant introduces a data-agnostic quantization method for diffusion transformers that eliminates the need for recalibration across timesteps and modalities, achieving state-of-the-art post-training quantization at low-bit settings for models like FLUX.1 and CogVideoX.
View Cached Full Text
Cached at: 07/06/26, 10:36 AM
Paper page - OrbitQuant: Data-Agnostic Quantization for Image and Video Diffusion Transformers
Source: https://huggingface.co/papers/2607.02461
Abstract
OrbitQuant enables efficient post-training quantization for diffusion transformers by using a normalized rotated basis that eliminates the need for recalibration across different timesteps and modalities.
Diffusion transformers(DiTs) achieve state-of-the-art image andvideo generation, but their multi-step sampling and growing parameter count make inference expensive.Post-training quantization(PTQ) is the natural remedy, yet DiT activations shift across timesteps, prompts, and guidance branches, forcing prior methods to re-fit calibration data for every new checkpoint or modality. We present OrbitQuant, a data-agnosticweight-activation quantizerthat bypasses range estimation by quantizing in a normalized, rotated basis. In this basis, arandomized permuted block-Hadamard(RPBH) rotation concentrates each coordinate around one fixed, known marginal regardless of the input, so a singleLloyd-Max codebookserves all timesteps, prompts, and layers of a given input dimension. We extend the same quantizer to weight rows offline, absorbing the rotation into the weights so that it cancels inside each linear layer and only a forward rotation on the activations remains at runtime. The same recipe transfers from image to video with no per-modality tuning. Across FLUX.1, Z-Image-Turbo, Wan 2.1, and CogVideoX, it sets the state of the art for PTQ at several low-bit settings. It also pushes PTQ of imagediffusion transformersto W2A4 with usable generation quality.
View arXiv pageView PDFProject pageAdd to collection
Get this paper in your agent:
hf papers read 2607\.02461
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2607.02461 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2607.02461 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2607.02461 in a Space README.md to link it from this page.
Collections including this paper1
Similar Articles
FourTune: Towards Fully 4-Bit Efficient Post-Training for Diffusion Models
FourTune proposes a fully 4-bit quantization framework (W4A4G4) for efficient post-training of diffusion models, using a triple-branch hybrid pipeline and custom fused kernels to reduce memory by 2.25× and increase throughput by 2.27× on 12B FLUX.1-dev without quality loss.
Theory-optimal Quantization Based on Flatness
Introduces Flatness metric and Bidirectional Diagonal Quantization (BDQ) for post-training quantization of large language models, achieving near-lossless 4-bit weight and activation quantization and substantial improvements at extreme low-bit settings.
Quantized Keys Steal Attention: Bias Correction for KV-Cache Compression in Video Diffusion
This paper identifies a bias in attention weights caused by quantizing keys in KV-cache compression for chunk-wise autoregressive video diffusion, and proposes a per-attention-score correction that removes the bias with negligible overhead, recovering near-BF16 video quality at INT2 quantization.
LoopQ: Quantization for Recursive Transformers
LoopQ is a post-training quantization framework for looped language models that addresses distribution shift, state reuse, and error accumulation. It achieves 68.8% average accuracy improvement under 4-bit weights and activations.
RateQuant: Optimal Mixed-Precision KV Cache Quantization via Rate-Distortion Theory
This paper introduces RateQuant, a method for optimal mixed-precision KV cache quantization that uses rate-distortion theory to address distortion model mismatch. It significantly reduces perplexity compared to existing methods like KIVI and QuaRot with minimal calibration overhead.