Skip to content

Sol-H3: Recursive Self-Improvement for MiniMax-H3 Inference Acceleration on Sol-Engine across Cloud and Edge

Sep 2026 · arXiv.org · Vol abs/2609.35110 · 4 citations · 83 references
Computer Science

TL;DR

A cross-resolution two-stage generation scheduler is introduced that exploits the step-wise nature of diffusion, completely eliminating the computationally expensive VAE decode-reencode cycle for resolution transferring cross different resolutions.

Abstract

Video diffusion models are rapidly scaling and exhibiting enhanced generation capabilities. Among these recent advancements, MiniMax-H3 stands out as a highly capable, production-level open-source model. However, its 33-billion parameters and multi-step iterative denoising process introduce substantial computational overhead. Consequently, their practical production is hindered by generation latency in the cloud deployment like NVIDIA-GB200, alongside strict memory limits that pose further challenges at the edge device like DGX-Spark. To address these diverse hardware bottlenecks from cloud to edge device, we present a full-stack inference pipeline that integrates efficient algorithmic design with optimized operator implementations. Algorithmically, we introduce a cross-resolution two-stage generation scheduler that exploits the step-wise nature of diffusion: early low-resolution steps rapidly establish the global layout, while later high-resolution steps focus refinements of local and perceptual details. These stages are connected by a learned latent-to-latent mapping module, completely eliminating the computationally expensive VAE decode-reencode cycle for resolution transferring cross different resolutions. For operator implementation, we deploy a Recursive Self-Improvement (RSI) loop that searches kernel fusions and memory layouts, evaluating latency together with numerical agreement. Together, these optimizations deliver up to 30x end-to-end speedup and 20% lower memory: a 5-second 1344x768 video with audio is generated 3.5x faster than real time on an 8xGB200 node, and in under a minute fully memory-resident on a single DGX Spark.

View source

Similar papers

#artificial intelligence Preprint Oct 2026

HGA-Quant: Heterogeneity-Aware Quantization for AutoRegressive Visual Generation Models

A training-free PTQ framework for ARVG, consisting of Tail-Aware Bounded-init Grid Refinement (T-BGR) and Coarse-to-Fine Activation Precision Scheduling (CAPS), which achieves near full-precision image generation quality at an effective W4A5.6 activation precision.

Xin-Meng Liu, Xiang-Long Yan, Kai-Cheng Yang et al. · 0 citations
Preprint Oct 2026

Two Halves are More than One: Phase-wise Velocity Distillation for Fast and High-Quality Image Generation

Recent diffusion-based image generation backbones have grown substantially in scale, making the network inference cost increase rapidly. While diffusion distillation techniques can reduce the number of inference steps, high-quality image generation within a single full-backbone-forward compute budget remains challengin...

Zhen-Yu Guo, Rong-Yuan Wu, Qiao-Si Yi et al. · 0 citations
Preprint Sep 2026

SparkDiffusion: Mitigating the High-Sparsity Trap --- A Unified Framework for up to $265\times$ Single-GPU Acceleration of Visual Generation

This work identifies the high-sparsity trap, a unified acceleration framework for visual generation that combines a short sparse warm-up, few-step trajectory-mixed distillation, and FP8 quantization with fused kernels, and sustains attention sparsity with strong visual quality on long-sequence 720P generation across Wa...

Yu-Xi Liu, Hao-Yu Li, Ze-Kun Zhang et al. · 0 citations
#computer vision Preprint Oct 2026

SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching

Diffusion-based world models enable high-quality interactive environment generation but suffer from substantial inference overhead due to repeated Transformer evaluations during denoising. Existing caching methods mainly exploit temporal redundancy at the feature or token level, leaving the underlying mathematical stru...

Zhen-Dong Mi, Pu Zhao, Zi-Yu Hu et al. · 0 citations
Preprint Oct 2026

Glance Forcing: Making Bidirectional Diffusion Autoregressive at a Glance

Glance Forcing is proposed, a lightweight distillation framework that decouples the denoising trajectory into two complementary slow and fast experts, each guided by a tailored, stage-specific objective, while keeping the pretrained teacher entirely frozen.

Zhuo-Bai Dong, Jun-Chao Yi, Hu-Jian Guo et al. · 0 citations
Preprint Oct 2026

LightVAE: Towards Compact and Efficient Video Autoencoders

Latent diffusion models enable high-quality video generation, but video VAE encoding and decoding remain computationally expensive. Existing methods use stage-specific lightweight operators, yet either constrain spatial and temporal computation jointly or remove temporal computation altogether, without distinguishing t...

Shi-Qiao Gu, Yang Yong, Feng-Jun Zhong et al. · 1 citation

Related blog posts

MIT News · Artificial Intelligence Oct 28, 2026

Making machines that can listen

Professor Paris Smaragdis develops computers with superhuman hearing for applications like speech recognition, music understanding, and environmental monitoring.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.