updata lightx2v

docs/PAPERS_ZH_CN/source/index.rst

+欢迎了解 Lightx2v 论文收藏集!
+==================
+
+.. figure:: ../../../assets/img_lightx2v.png
+  :width: 80%
+  :align: center
+  :alt: Lightx2v
+  :class: no-scaled-link
+
+.. raw:: html
+
+    <div align="center" style="font-family: charter;">
+
+    <a href="https://opensource.org/licenses/Apache-2.0"><img src="https://img.shields.io/badge/License-Apache_2.0-blue.svg" alt="License"></a>
+    <a href="https://deepwiki.com/ModelTC/lightx2v"><img src="https://deepwiki.com/badge.svg" alt="Ask DeepWiki"></a>
+    <a href="https://lightx2v-en.readthedocs.io/en/latest"><img src="https://img.shields.io/badge/docs-English-99cc2" alt="Doc"></a>
+    <a href="https://lightx2v-zhcn.readthedocs.io/zh-cn/latest"><img src="https://img.shields.io/badge/文档-中文-99cc2" alt="Doc"></a>
+    <a href="https://lightx2v-papers-zhcn.readthedocs.io/zh-cn/latest"><img src="https://img.shields.io/badge/论文集-中文-99cc2" alt="Papers"></a>
+    <a href="https://hub.docker.com/r/lightx2v/lightx2v/tags"><img src="https://badgen.net/badge/icon/docker?icon=docker&label" alt="Docker"></a>
+
+    </div>
+
+    <div align="center" style="font-family: charter;">
+    <strong>LightX2V: 一个轻量级的视频生成推理框架</strong>
+    </div>
+
+
+LightX2V 是一个轻量级的视频生成推理框架。这里是我们维护的一个视频生成推理加速相关的论文收藏集，帮助你快速了解视频生成推理加速相关的经典方法和最新进展。
+
+GitHub: https://github.com/ModelTC/lightx2v
+
+HuggingFace: https://huggingface.co/lightx2v
+
+论文收藏集
+-------------
+
+.. toctree::
+   :maxdepth: 1
+   :caption: 论文分类
+
+   图像视频生成基础 <papers/generation_basics.md>
+   开源模型 <papers/models.md>
+   模型量化 <papers/quantization.md>
+   特征缓存 <papers/cache.md>
+   注意力机制 <papers/attention.md>
+   参数卸载 <papers/offload.md>
+   并行推理 <papers/parallel.md>
+   变分辨率推理 <papers/changing_resolution.md>
+   步数蒸馏 <papers/step_distill.md>
+   自回归模型 <papers/autoregressive.md>
+   vae加速 <papers/vae.md>
+   prompt增强 <papers/prompt_enhance.md>
+   强化学习 <papers/RL.md>

docs/PAPERS_ZH_CN/source/papers/RL.md

+# 强化学习
+
+xxx

docs/PAPERS_ZH_CN/source/papers/attention.md

+# 注意力机制
+
+### Sparse VideoGen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity
+
+[paper](https://arxiv.org/abs/2502.01776) | [code](https://github.com/svg-project/Sparse-VideoGen)
+
+### Sparse VideoGen2: Accelerate Video Generation with Sparse Attention via Semantic-Aware Permutation
+
+[paper](https://arxiv.org/abs/2505.18875)
+
+### Training-free and Adaptive Sparse Attention for Efficient Long Video Generation
+
+[paper](https://arxiv.org/abs/2502.21079)
+
+### DSV: Exploiting Dynamic Sparsity to Accelerate Large-Scale Video DiT Training
+
+[paper](https://arxiv.org/abs/2502.07590)
+
+### MMInference: Accelerating Pre-filling for Long-Context VLMs via Modality-Aware Permutation Sparse Attention
+
+[paper](https://github.com/microsoft/MInference)
+
+### FPSAttention: Training-Aware FP8 and Sparsity Co-Design for Fast Video Diffusion
+
+[paper](https://arxiv.org/abs/2506.04648)
+
+### VORTA: Efficient Video Diffusion via Routing Sparse Attention
+
+[paper](https://arxiv.org/abs/2505.18809)
+
+### Training-Free Efficient Video Generation via Dynamic Token Carving
+
+[paper](https://arxiv.org/abs/2505.16864)
+
+### RainFusion: Adaptive Video Generation Acceleration via Multi-Dimensional Visual Redundancy
+
+[paper](https://arxiv.org/abs/2505.21036)
+
+### Radial Attention: O(nlogn) Sparse Attention with Energy Decay for Long Video Generation
+
+[paper](https://arxiv.org/abs/2506.19852)
+
+### VMoBA: Mixture-of-Block Attention for Video Diffusion Models
+
+[paper](https://arxiv.org/abs/2506.23858)
+
+### SpargeAttention: Accurate and Training-free Sparse Attention Accelerating Any Model Inference
+
+[paper](https://arxiv.org/abs/2502.18137) | [code](https://github.com/thu-ml/SpargeAttn)
+
+### Fast Video Generation with Sliding Tile Attention
+
+[paper](https://arxiv.org/abs/2502.04507) | [code](https://github.com/hao-ai-lab/FastVideo)
+
+### PAROAttention: Pattern-Aware ReOrdering for Efficient Sparse and Quantized Attention in Visual Generation Models
+
+[paper](https://arxiv.org/abs/2506.16054)
+
+### Generalized Neighborhood Attention: Multi-dimensional Sparse Attention at the Speed of Light
+
+[paper](https://arxiv.org/abs/2504.16922)
+
+### Astraea: A GPU-Oriented Token-wise Acceleration Framework for Video Diffusion Transformers
+
+[paper](https://arxiv.org/abs/2506.05096)
+
+### ∇NABLA: Neighborhood Adaptive Block-Level Attention
+
+[paper](https://arxiv.org/abs/2507.13546v1) [code](https://github.com/gen-ai-team/Wan2.1-NABLA)
+
+### Compact Attention: Exploiting Structured Spatio-Temporal Sparsity for Fast Video Generation
+
+[paper](https://arxiv.org/abs/2508.12969)
+
+### A Survey of Efficient Attention Methods: Hardware-efficient, Sparse, Compact, and Linear Attention
+
+[paper](https://attention-survey.github.io/files/Attention_Survey.pdf)
+
+### Bidirectional Sparse Attention for Faster Video Diffusion Training
+
+[paper](https://arxiv.org/abs/2509.01085)
+
+### Mixture of Contexts for Long Video Generation
+
+[paper](https://arxiv.org/abs/2508.21058)
+
+### LoViC: Efficient Long Video Generation with Context Compression
+
+[paper](https://arxiv.org/abs/2507.12952)
+
+### MagiAttention: A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Mask Training
+
+[paper](https://sandai-org.github.io/MagiAttention/blog/) [code](https://github.com/SandAI-org/MagiAttention)
+
+### DraftAttention: Fast Video Diffusion via Low-Resolution Attention Guidance
+
+[paper](https://arxiv.org/abs/2505.14708) [code](https://github.com/shawnricecake/draft-attention)
+
+### XAttention: Block Sparse Attention with Antidiagonal Scoring
+
+[paper](https://arxiv.org/abs/2503.16428) [code](https://github.com/mit-han-lab/x-attention)
+
+### VSA: Faster Video Diffusion with Trainable Sparse Attention
+
+[paper](https://arxiv.org/abs/2505.13389) [code](https://github.com/hao-ai-lab/FastVideo)
+
+### QuantSparse: Comprehensively Compressing Video Diffusion Transformer with Model Quantization and Attention Sparsification
+
+[paper](https://arxiv.org/abs/2509.23681)
+
+### SLA: Beyond Sparsity in Diffusion Transformers via Fine-Tunable Sparse-Linear Attention
+
+[paper](https://arxiv.org/abs/2509.24006)

docs/PAPERS_ZH_CN/source/papers/autoregressive.md

+# 自回归模型
+
+xxx

docs/PAPERS_ZH_CN/source/papers/cache.md

+# 特征缓存
+
+xxx

docs/PAPERS_ZH_CN/source/papers/changing_resolution.md

+# 变分辨率推理
+
+xxx

docs/PAPERS_ZH_CN/source/papers/generation_basics.md

+# 图像视频生成基础
+
+xxx

docs/PAPERS_ZH_CN/source/papers/models.md

+
+<div align=center>
+
+# Open-Source Models
+
+📢: Collections of Awesome Open-Source Model Resources.
+
+</div>
+
+
+## 📚 <span id="head1"> *Contents* </span>
+
+- Open-Source Models
+  - [Foundation Models](#foundation-models)
+  - [World Models](#world-models)
+
+
+### Foundation Models:
+
+- **Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets**, Technical Report 2023.
+
+  *Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, et al.*
+
+  [[Paper](https://arxiv.org/abs/2311.15127)] [[Code](https://github.com/Stability-AI/generative-models)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/UNet-brown)
+
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{blattmann2023stable,
+    title={Stable video diffusion: Scaling latent video diffusion models to large datasets},
+    author={Blattmann, Andreas and Dockhorn, Tim and Kulal, Sumith and Mendelevitch, Daniel and Kilian, Maciej and Lorenz, Dominik and Levi, Yam and English, Zion and Voleti, Vikram and Letts, Adam and others},
+    journal={arXiv preprint arXiv:2311.15127},
+    year={2023}
+  }
+  ```
+  </details>
+
+- **Wan: Open and Advanced Large-Scale Video Generative Models**, Technical Report 2025.
+
+  *Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, et al.*
+
+  [[Paper](https://arxiv.org/abs/2503.20314)] [[Code](https://github.com/Wan-Video/Wan2.1)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/DIT-brown)
+
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{wan2025wan,
+  title={Wan: Open and advanced large-scale video generative models},
+  author={Wan, Team and Wang, Ang and Ai, Baole and Wen, Bin and Mao, Chaojie and Xie, Chen-Wei and Chen, Di and Yu, Feiwu and Zhao, Haiming and Yang, Jianxiao and others},
+  journal={arXiv preprint arXiv:2503.20314},
+  year={2025}
+  }
+  ```
+  </details>
+
+- **HunyuanVideo: A Systematic Framework For Large Video Generation Model**, Technical Report 2024.
+
+  *Weijie Kong, Qi Tian, Zijian Zhang, Rox Min, Zuozhuo Dai, Jin Zhou, et al.*
+
+  [[Paper](https://arxiv.org/abs/2412.03603)] [[Code](https://github.com/Tencent-Hunyuan/HunyuanVideo)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/DIT-brown)
+
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{kong2024hunyuanvideo,
+  title={Hunyuanvideo: A systematic framework for large video generative models},
+  author={Kong, Weijie and Tian, Qi and Zhang, Zijian and Min, Rox and Dai, Zuozhuo and Zhou, Jin and Xiong, Jiangfeng and Li, Xin and Wu, Bo and Zhang, Jianwei and others},
+  journal={arXiv preprint arXiv:2412.03603},
+  year={2024}
+  }
+  ```
+  </details>
+
+- **CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer**, ICLR 2025.
+
+  *Zhuoyi Yang, Jiayan Teng, Wendi Zheng, Ming Ding, Shiyu Huang, Jiazheng Xu, et al.*
+
+  [[Paper](https://arxiv.org/abs/2408.06072)] [[Code](https://github.com/zai-org/CogVideo)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/DIT-brown)
+
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{yang2024cogvideox,
+    title={Cogvideox: Text-to-video diffusion models with an expert transformer},
+    author={Yang, Zhuoyi and Teng, Jiayan and Zheng, Wendi and Ding, Ming and Huang, Shiyu and Xu, Jiazheng and Yang, Yuanming and Hong, Wenyi and Zhang, Xiaohan and Feng, Guanyu and others},
+    journal={arXiv preprint arXiv:2408.06072},
+    year={2024}
+  }
+  ```
+  </details>
+
+
+- **SkyReels V2: Infinite-Length Film Generative Model**, Technical Report 2025.
+
+  *Guibin Chen, Dixuan Lin, Jiangping Yang, Chunze Lin, Junchen Zhu, Mingyuan Fan, et al.*
+
+  [[Paper](https://arxiv.org/abs/2504.13074)] [[Code](https://github.com/SkyworkAI/SkyReels-V2)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/DIT-brown)
+
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @misc{chen2025skyreelsv2infinitelengthfilmgenerative,
+        title={SkyReels-V2: Infinite-length Film Generative Model},
+        author={Guibin Chen and Dixuan Lin and Jiangping Yang and Chunze Lin and Junchen Zhu and Mingyuan Fan and Hao Zhang and Sheng Chen and Zheng Chen and Chengcheng Ma and Weiming Xiong and Wei Wang and Nuo Pang and Kang Kang and Zhiheng Xu and Yuzhe Jin and Yupeng Liang and Yubing Song and Peng Zhao and Boyuan Xu and Di Qiu and Debang Li and Zhengcong Fei and Yang Li and Yahui Zhou},
+        year={2025},
+        eprint={2504.13074},
+        archivePrefix={arXiv},
+        primaryClass={cs.CV},
+        url={https://arxiv.org/abs/2504.13074},
+  }
+  ```
+  </details>
+
+
+- **Open-Sora: Democratizing Efficient Video Production for All**, Technical Report 2025.
+
+  *Xiangyu Peng, Zangwei Zheng, Chenhui Shen, Tom Young, Xinying Guo, et al.*
+
+  [[Paper](https://arxiv.org/abs/2503.09642v2)] [[Code](https://github.com/hpcaitech/Open-Sora)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/DIT-brown) ![](https://img.shields.io/badge/V2V-orange)
+
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{peng2025open,
+    title={Open-sora 2.0: Training a commercial-level video generation model in $200 k},
+    author={Peng, Xiangyu and Zheng, Zangwei and Shen, Chenhui and Young, Tom and Guo, Xinying and Wang, Binluo and Xu, Hang and Liu, Hongxin and Jiang, Mingyan and Li, Wenjun and others},
+    journal={arXiv preprint arXiv:2503.09642},
+    year={2025}
+  }
+  ```
+  </details>
+
+- **Pyramidal Flow Matching for Efficient Video Generative Modeling**, Technical Report 2024.
+
+  *Yang Jin, Zhicheng Sun, Ningyuan Li, Kun Xu, Kun Xu, et al.*
+
+  [[Paper](https://arxiv.org/abs/2410.05954)] [[Code](https://github.com/jy0205/Pyramid-Flow)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/AR-brown)
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{jin2024pyramidal,
+  title={Pyramidal flow matching for efficient video generative modeling},
+  author={Jin, Yang and Sun, Zhicheng and Li, Ningyuan and Xu, Kun and Jiang, Hao and Zhuang, Nan and Huang, Quzhe and Song, Yang and Mu, Yadong and Lin, Zhouchen},
+  journal={arXiv preprint arXiv:2410.05954},
+  year={2024}
+  }
+  ```
+  </details>
+
+- **MAGI-1: Autoregressive Video Generation at Scale**, Technical Report 2025.
+
+  *Sand.ai, Hansi Teng, Hongyu Jia, Lei Sun, Lingzhi Li, Maolin Li, Mingqiu Tang, et al.*
+
+  [[Paper](https://arxiv.org/pdf/2505.13211)] [[Code](https://github.com/SandAI-org/Magi-1)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/AR-brown) ![](https://img.shields.io/badge/V2V-orange)
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{teng2025magi,
+    title={MAGI-1: Autoregressive Video Generation at Scale},
+    author={Teng, Hansi and Jia, Hongyu and Sun, Lei and Li, Lingzhi and Li, Maolin and Tang, Mingqiu and Han, Shuai and Zhang, Tianning and Zhang, WQ and Luo, Weifeng and others},
+    journal={arXiv preprint arXiv:2505.13211},
+    year={2025}
+  }
+  ```
+  </details>
+
+- **From Slow Bidirectional to Fast Autoregressive Video Diffusion Models**, CVPR 2025.
+
+  *Tianwei Yin, Qiang Zhang, Richard Zhang, William T. Freeman, Fredo Durand, et al.*
+
+  [[Paper](http://arxiv.org/abs/2412.07772)] [[Code](https://github.com/tianweiy/CausVid)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/AR-brown)
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @inproceedings{yin2025slow,
+  title={From slow bidirectional to fast autoregressive video diffusion models},
+  author={Yin, Tianwei and Zhang, Qiang and Zhang, Richard and Freeman, William T and Durand, Fredo and Shechtman, Eli and Huang, Xun},
+  booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
+  pages={22963--22974},
+  year={2025}
+  }
+  ```
+  </details>
+
+- **Packing Input Frame Context in Next-Frame Prediction Models for Video Generation**, arxiv 2025.
+
+  *Lvmin Zhang, Maneesh Agrawala.*
+
+  [[Paper](https://arxiv.org/abs/2504.12626)] [[Code](https://github.com/lllyasviel/FramePack)] ![](https://img.shields.io/badge/T2V-blue) ![](https://img.shields.io/badge/I2V-green) ![](https://img.shields.io/badge/AR-brown)
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{zhang2025packing,
+    title={Packing input frame context in next-frame prediction models for video generation},
+    author={Zhang, Lvmin and Agrawala, Maneesh},
+    journal={arXiv preprint arXiv:2504.12626},
+    year={2025}
+  }
+  ```
+  </details>
+
+### World Models:
+
+- **Matrix-Game 2.0: An Open-Source, Real-Time, and Streaming Interactive World Model**, Technical Report 2025.
+
+  *Xianglong He, Chunli Peng, Zexiang Liu, Boyang Wang, Yifan Zhang, et al.*
+
+  [[Paper](https://arxiv.org/abs/2508.13009)] [[Code](https://matrix-game-v2.github.io/)] ![](https://img.shields.io/badge/keyboard-blue) ![](https://img.shields.io/badge/mouse-green) ![](https://img.shields.io/badge/DIT-brown)
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{he2025matrix,
+    title={Matrix-Game 2.0: An Open-Source, Real-Time, and Streaming Interactive World Model},
+    author={He, Xianglong and Peng, Chunli and Liu, Zexiang and Wang, Boyang and Zhang, Yifan and Cui, Qi and Kang, Fei and Jiang, Biao and An, Mengyin and Ren, Yangyang and others},
+    journal={arXiv preprint arXiv:2508.13009},
+    year={2025}
+  }
+  ```
+  </details>
+
+- **HunyuanWorld 1.0: Generating Immersive, Explorable, and Interactive 3D Worlds from Words or Pixels**, Technical Report 2025.
+
+  *HunyuanWorld Team, Zhenwei Wang, Yuhao Liu, Junta Wu, Zixiao Gu, Haoyuan Wang, et al.*
+
+  [[Paper](https://arxiv.org/abs/2507.21809)] [[Code](https://github.com/Tencent-Hunyuan/HunyuanWorld-1.0)] ![](https://img.shields.io/badge/image-blue) ![](https://img.shields.io/badge/text-green) ![](https://img.shields.io/badge/DIT-brown)
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{team2025hunyuanworld,
+    title={HunyuanWorld 1.0: Generating Immersive, Explorable, and Interactive 3D Worlds from Words or Pixels},
+    author={Team, HunyuanWorld and Wang, Zhenwei and Liu, Yuhao and Wu, Junta and Gu, Zixiao and Wang, Haoyuan and Zuo, Xuhui and Huang, Tianyu and Li, Wenhuan and Zhang, Sheng and others},
+    journal={arXiv preprint arXiv:2507.21809},
+    year={2025}
+  }
+  ```
+  </details>
+
+- **Cosmos-Drive-Dreams: Scalable Synthetic Driving Data Generation with World Foundation Models**, Technical Report 2025.
+
+  *Xuanchi Ren, Yifan Lu, Tianshi Cao, Ruiyuan Gao, Shengyu Huang, Amirmojtaba Sabour, et al.*
+
+  [[Paper](https://arxiv.org/abs/2506.09042)] [[Code](https://research.nvidia.com/labs/toronto-ai/cosmos_drive_dreams)] ![](https://img.shields.io/badge/drive-blue) ![](https://img.shields.io/badge/DIT-brown)
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{ren2025cosmos,
+    title={Cosmos-Drive-Dreams: Scalable Synthetic Driving Data Generation with World Foundation Models},
+    author={Ren, Xuanchi and Lu, Yifan and Cao, Tianshi and Gao, Ruiyuan and Huang, Shengyu and Sabour, Amirmojtaba and Shen, Tianchang and Pfaff, Tobias and Wu, Jay Zhangjie and Chen, Runjian and others},
+    journal={arXiv preprint arXiv:2506.09042},
+    year={2025}
+  }
+  ```
+  </details>
+
+- **Genie 3: A new frontier for world models**, Blog 2025.
+
+  *Google DeepMind*
+
+  [[Blog](https://deepmind.google/discover/blog/genie-3-a-new-frontier-for-world-models/)] ![](https://img.shields.io/badge/event-blue) ![](https://img.shields.io/badge/DIT-brown)
+
+- **GAIA-2: A Controllable Multi-View Generative World Model for Autonomous Driving.**, Technical Report 2025.
+
+  *Lloyd Russell, Anthony Hu, Lorenzo Bertoni, George Fedoseev, Jamie Shotton, et al.*
+
+  [[Paper](https://arxiv.org/abs/2503.20523)] [[Code](https://github.com/Tencent-Hunyuan/HunyuanWorld-1.0)] ![](https://img.shields.io/badge/drive-blue) ![](https://img.shields.io/badge/transformer-brown)
+  <details> <summary>BibTex</summary>
+
+  ```text
+  @article{russell2025gaia,
+    title={Gaia-2: A controllable multi-view generative world model for autonomous driving},
+    author={Russell, Lloyd and Hu, Anthony and Bertoni, Lorenzo and Fedoseev, George and Shotton, Jamie and Arani, Elahe and Corrado, Gianluca},
+    journal={arXiv preprint arXiv:2503.20523},
+    year={2025}
+  }
+  ```
+  </details>

docs/PAPERS_ZH_CN/source/papers/offload.md

+# 参数卸载
+
+xxx

docs/PAPERS_ZH_CN/source/papers/parallel.md

+# 并行推理
+
+xxx

docs/PAPERS_ZH_CN/source/papers/prompt_enhance.md

+# prompt增强
+
+xxx

docs/PAPERS_ZH_CN/source/papers/quantization.md

+# 模型量化
+
+xxx

docs/PAPERS_ZH_CN/source/papers/step_distill.md

+# 步数蒸馏
+
+xxx

docs/PAPERS_ZH_CN/source/papers/vae.md

+# vae加速
+
+xxx

docs/ZH_CN/.readthedocs.yaml

+version: 2
+
+# Set the version of Python and other tools you might need
+build:
+  os: ubuntu-20.04
+  tools:
+    python: "3.10"
+
+formats:
+    - epub
+
+sphinx:
+  configuration: docs/ZH_CN/source/conf.py
+
+python:
+  install:
+    - requirements: requirements-docs.txt

docs/ZH_CN/Makefile

+# Minimal makefile for Sphinx documentation
+#
+
+# You can set these variables from the command line, and also
+# from the environment for the first two.
+SPHINXOPTS    ?=
+SPHINXBUILD   ?= sphinx-build
+SOURCEDIR     = source
+BUILDDIR      = build
+
+# Put it first so that "make" without argument is like "make help".
+help:
+	@$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+.PHONY: help Makefile
+
+# Catch-all target: route all unknown targets to Sphinx using the new
+# "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
+%: Makefile
+	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)

docs/ZH_CN/make.bat

+@ECHO OFF
+
+pushd %~dp0
+
+REM Command file for Sphinx documentation
+
+if "%SPHINXBUILD%" == "" (
+	set SPHINXBUILD=sphinx-build
+)
+set SOURCEDIR=source
+set BUILDDIR=build
+
+%SPHINXBUILD% >NUL 2>NUL
+if errorlevel 9009 (
+	echo.
+	echo.The 'sphinx-build' command was not found. Make sure you have Sphinx
+	echo.installed, then set the SPHINXBUILD environment variable to point
+	echo.to the full path of the 'sphinx-build' executable. Alternatively you
+	echo.may add the Sphinx directory to PATH.
+	echo.
+	echo.If you don't have Sphinx installed, grab it from
+	echo.https://www.sphinx-doc.org/
+	exit /b 1
+)
+
+if "%1" == "" goto help
+
+%SPHINXBUILD% -M %1 %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% %O%
+goto end
+
+:help
+%SPHINXBUILD% -M help %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% %O%
+
+:end
+popd

docs/ZH_CN/source/conf.py

+# Configuration file for the Sphinx documentation builder.
+#
+# This file only contains a selection of the most common options. For a full
+# list see the documentation:
+# https://www.sphinx-doc.org/en/master/usage/configuration.html
+
+# -- Path setup --------------------------------------------------------------
+
+# If extensions (or modules to document with autodoc) are in another directory,
+# add these directories to sys.path here. If the directory is relative to the
+# documentation root, use os.path.abspath to make it absolute, like shown here.
+
+import logging
+import os
+import sys
+from typing import List
+
+import sphinxcontrib.redoc
+from sphinx.ext import autodoc
+
+logger = logging.getLogger(__name__)
+sys.path.append(os.path.abspath("../.."))
+
+# -- Project information -----------------------------------------------------
+
+project = "Lightx2v"
+copyright = "2025, Lightx2v Team"
+author = "the Lightx2v Team"
+
+# -- General configuration ---------------------------------------------------
+
+# Add any Sphinx extension module names here, as strings. They can be
+# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
+# ones.
+extensions = [
+    "sphinx.ext.napoleon",
+    "sphinx.ext.viewcode",
+    "sphinx.ext.intersphinx",
+    "sphinx_copybutton",
+    "sphinx.ext.autodoc",
+    "sphinx.ext.autosummary",
+    "sphinx.ext.mathjax",
+    "myst_parser",
+    "sphinxarg.ext",
+    "sphinxcontrib.redoc",
+    "sphinxcontrib.openapi",
+]
+
+myst_enable_extensions = [
+    "dollarmath",
+    "amsmath",
+]
+
+html_static_path = ["_static"]
+
+# Add any paths that contain templates here, relative to this directory.
+templates_path = ["_templates"]
+
+# List of patterns, relative to source directory, that match files and
+# directories to ignore when looking for source files.
+# This pattern also affects html_static_path and html_extra_path.
+exclude_patterns: List[str] = ["**/*.template.rst"]
+
+# Exclude the prompt "$" when copying code
+copybutton_prompt_text = r"\$ "
+copybutton_prompt_is_regexp = True
+
+# -- Options for HTML output -------------------------------------------------
+
+# The theme to use for HTML and HTML Help pages.  See the documentation for
+# a list of builtin themes.
+#
+html_title = project
+html_theme = "sphinx_book_theme"
+# html_theme = 'sphinx_rtd_theme'
+html_logo = "../../../assets/img_lightx2v.png"
+html_theme_options = {
+    "path_to_docs": "docs/ZH_CN/source",
+    "repository_url": "https://github.com/ModelTC/lightx2v",
+    "use_repository_button": True,
+}
+
+# Add any paths that contain custom static files (such as style sheets) here,
+# relative to this directory. They are copied after the builtin static files,
+# so a file named "default.css" will overwrite the builtin "default.css".
+# html_static_path = ['_static']
+
+
+# Generate additional rst documentation here.
+def setup(app):
+    # from docs.source.generate_examples import generate_examples
+    # generate_examples()
+    pass
+
+
+# Mock out external dependencies here.
+autodoc_mock_imports = [
+    "cpuinfo",
+    "torch",
+    "transformers",
+    "psutil",
+    "prometheus_client",
+    "sentencepiece",
+    "lightllmnumpy",
+    "tqdm",
+    "tensorizer",
+]
+
+for mock_target in autodoc_mock_imports:
+    if mock_target in sys.modules:
+        logger.info(
+            "Potentially problematic mock target (%s) found; autodoc_mock_imports cannot mock modules that have already been loaded into sys.modules when the sphinx build starts.",
+            mock_target,
+        )
+
+
+class MockedClassDocumenter(autodoc.ClassDocumenter):
+    """Remove note about base class when a class is derived from object."""
+
+    def add_line(self, line: str, source: str, *lineno: int) -> None:
+        if line == "   Bases: :py:class:`object`":
+            return
+        super().add_line(line, source, *lineno)
+
+
+autodoc.ClassDocumenter = MockedClassDocumenter
+
+navigation_with_keys = False

docs/ZH_CN/source/deploy_guides/deploy_comfyui.md

+# ComfyUI 部署
+
+## ComfyUI-Lightx2vWrapper
+
+LightX2V 的官方 ComfyUI 集成节点已经发布在独立仓库中，提供了完整的模块化配置系统和优化功能。
+
+### 项目地址
+
+- GitHub: [https://github.com/ModelTC/ComfyUI-Lightx2vWrapper](https://github.com/ModelTC/ComfyUI-Lightx2vWrapper)
+
+### 主要特性
+
+- 模块化配置系统：为视频生成的各个方面提供独立节点
+- 支持文生视频（T2V）和图生视频（I2V）两种生成模式
+- 高级优化功能：
+  - TeaCache 加速（最高 3 倍加速）
+  - 量化支持（int8、fp8）
+  - CPU 卸载内存优化
+  - 轻量级 VAE 选项
+- LoRA 支持：可链式组合多个 LoRA 模型
+- 多模型支持：wan2.1、hunyuan 等架构
+
+### 安装和使用
+
+请访问上述 GitHub 仓库查看详细的安装说明、使用教程和示例工作流。

docs/ZH_CN/source/deploy_guides/deploy_gradio.md

+# Gradio 部署指南
+
+## 📖 概述
+
+Lightx2v 是一个轻量级的视频推理和生成引擎，提供基于 Gradio 的 Web 界面，支持图像到视频（Image-to-Video）和文本到视频（Text-to-Video）两种生成模式。
+
+对于Windows系统，我们提供了便捷的一键部署方式，支持自动环境配置和智能参数优化。详细操作请参考[一键启动Gradio](./deploy_local_windows.md/#一键启动gradio推荐)章节。
+
+![Gradio中文界面](../../../../assets/figs/portabl_windows/pic_gradio_zh.png)
+
+## 📁 文件结构
+
+```
+LightX2V/app/
+├── gradio_demo.py          # 英文界面演示
+├── gradio_demo_zh.py       # 中文界面演示
+├── run_gradio.sh          # 启动脚本
+├── README.md              # 说明文档
+├── outputs/               # 生成视频保存目录
+└── inference_logs.log     # 推理日志
+```
+
+本项目包含两个主要演示文件：
+- `gradio_demo.py` - 英文界面版本
+- `gradio_demo_zh.py` - 中文界面版本
+
+## 🚀 快速开始
+
+### 环境要求
+
+按照[快速开始文档](../getting_started/quickstart.md)安装环境
+
+#### 推荐优化库配置
+
+- ✅ [Flash attention](https://github.com/Dao-AILab/flash-attention)
+- ✅ [Sage attention](https://github.com/thu-ml/SageAttention)
+- ✅ [vllm-kernel](https://github.com/vllm-project/vllm)
+- ✅ [sglang-kernel](https://github.com/sgl-project/sglang/tree/main/sgl-kernel)
+- ✅ [q8-kernel](https://github.com/KONAKONA666/q8_kernels) (仅支持ADA架构的GPU)
+
+可根据需要，按照各算子的项目主页教程进行安装。
+
+### 📥 模型下载
+
+可参考[模型结构文档](../getting_started/model_structure.md)下载完整模型（包含量化和非量化版本）或仅下载量化/非量化版本。
+
+#### wan2.1 模型目录结构
+
+```
+models/
+├── wan2.1_i2v_720p_lightx2v_4step.safetensors                   # 原始精度
+├── wan2.1_i2v_720p_scaled_fp8_e4m3_lightx2v_4step.safetensors   # FP8 量化
+├── wan2.1_i2v_720p_int8_lightx2v_4step.safetensors              # INT8 量化
+├── wan2.1_i2v_720p_int8_lightx2v_4step_split                    # INT8 量化分block存储目录
+├── wan2.1_i2v_720p_scaled_fp8_e4m3_lightx2v_4step_split         # FP8 量化分block存储目录
+├── 其他权重(例如t2v)
+├── t5/clip/xlm-roberta-large/google    # text和image encoder
+├── vae/lightvae/lighttae               # vae
+└── config.json                         # 模型配置文件
+```
+
+#### wan2.2 模型目录结构
+
+```
+models/
+├── wan2.2_i2v_A14b_high_noise_lightx2v_4step_1030.safetensors        # high noise 原始精度
+├── wan2.2_i2v_A14b_high_noise_fp8_e4m3_lightx2v_4step_1030.safetensors    # high noise FP8 量化
+├── wan2.2_i2v_A14b_high_noise_int8_lightx2v_4step_1030.safetensors   # high noise INT8 量化
+├── wan2.2_i2v_A14b_high_noise_int8_lightx2v_4step_1030_split         # high noise INT8 量化分block存储目录
+├── wan2.2_i2v_A14b_low_noise_lightx2v_4step.safetensors         # low noise 原始精度
+├── wan2.2_i2v_A14b_low_noise_fp8_e4m3_lightx2v_4step.safetensors     # low noise FP8 量化
+├── wan2.2_i2v_A14b_low_noise_int8_lightx2v_4step.safetensors    # low noise INT8 量化
+├── wan2.2_i2v_A14b_low_noise_int8_lightx2v_4step_split          # low noise INT8 量化分block存储目录
+├── t5/clip/xlm-roberta-large/google    # text和image encoder
+├── vae/lightvae/lighttae               # vae
+└── config.json                         # 模型配置文件
+```
+
+**📝 下载说明**：
+
+- 模型权重可从 HuggingFace 下载：
+  - [Wan2.1-Distill-Models](https://huggingface.co/lightx2v/Wan2.1-Distill-Models)
+  - [Wan2.2-Distill-Models](https://huggingface.co/lightx2v/Wan2.2-Distill-Models)
+- Text 和 Image Encoder 可从 [Encoders](https://huggingface.co/lightx2v/Encoders) 下载
+- VAE 可从 [Autoencoders](https://huggingface.co/lightx2v/Autoencoders) 下载
+- 对于 `xxx_split` 目录（例如 `wan2.1_i2v_720p_scaled_fp8_e4m3_lightx2v_4step_split`），即按照 block 存储的多个 safetensors，适用于内存不足的设备。例如内存 16GB 以内，请根据自身情况下载
+
+
+### 启动方式
+
+#### 方式一：使用启动脚本（推荐）
+
+**Linux 环境：**
+```bash
+# 1. 编辑启动脚本，配置相关路径
+cd app/
+vim run_gradio.sh
+
+# 需要修改的配置项：
+# - lightx2v_path: Lightx2v项目根目录路径
+# - model_path: 模型根目录路径（包含所有模型文件）
+
+# 💾 重要提示：建议将模型路径指向SSD存储位置
+# 例如：/mnt/ssd/models/ 或 /data/ssd/models/
+
+# 2. 运行启动脚本
+bash run_gradio.sh
+
+# 3. 或使用参数启动
+bash run_gradio.sh --lang zh --port 8032
+bash run_gradio.sh --lang en --port 7862
+```
+
+**Windows 环境：**
+```cmd
+# 1. 编辑启动脚本，配置相关路径
+cd app\
+notepad run_gradio_win.bat
+
+# 需要修改的配置项：
+# - lightx2v_path: Lightx2v项目根目录路径
+# - model_path: 模型根目录路径（包含所有模型文件）
+
+# 💾 重要提示：建议将模型路径指向SSD存储位置
+# 例如：D:\models\ 或 E:\models\
+
+# 2. 运行启动脚本
+run_gradio_win.bat
+
+# 3. 或使用参数启动
+run_gradio_win.bat --lang zh --port 8032
+run_gradio_win.bat --lang en --port 7862
+```
+
+#### 方式二：直接命令行启动
+
+```bash
+pip install -v git+https://github.com/ModelTC/LightX2V.git
+```
+
+**Linux 环境：**
+
+**中文界面版本：**
+```bash
+python gradio_demo_zh.py \
+    --model_path /path/to/models \
+    --server_name 0.0.0.0 \
+    --server_port 7862
+```
+
+**英文界面版本：**
+```bash
+python gradio_demo.py \
+    --model_path /path/to/models \
+    --server_name 0.0.0.0 \
+    --server_port 7862
+```
+
+**Windows 环境：**
+
+**中文界面版本：**
+```cmd
+python gradio_demo_zh.py ^
+    --model_path D:\models ^
+    --server_name 127.0.0.1 ^
+    --server_port 7862
+```
+
+**英文界面版本：**
+```cmd
+python gradio_demo.py ^
+    --model_path D:\models ^
+    --server_name 127.0.0.1 ^
+    --server_port 7862
+```
+
+**💡 提示**：模型类型（wan2.1/wan2.2）、任务类型（i2v/t2v）以及具体的模型文件选择均在 Web 界面中进行配置。
+
+## 📋 命令行参数
+
+| 参数 | 类型 | 必需 | 默认值 | 说明 |
+|------|------|------|--------|------|
+| `--model_path` | str | ✅ | - | 模型根目录路径（包含所有模型文件的目录） |
+| `--server_port` | int | ❌ | 7862 | 服务器端口 |
+| `--server_name` | str | ❌ | 0.0.0.0 | 服务器IP地址 |
+| `--output_dir` | str | ❌ | ./outputs | 输出视频保存目录 |
+
+**💡 说明**：模型类型（wan2.1/wan2.2）、任务类型（i2v/t2v）以及具体的模型文件选择均在 Web 界面中进行配置。
+
+## 🎯 功能特性
+
+### 模型配置
+
+- **模型类型**: 支持 wan2.1 和 wan2.2 两种模型架构
+- **任务类型**: 支持图像到视频（i2v）和文本到视频（t2v）两种生成模式
+- **模型选择**: 前端自动识别并筛选可用的模型文件，支持自动检测量化精度
+- **编码器配置**: 支持选择 T5 文本编码器、CLIP 图像编码器和 VAE 解码器
+- **算子选择**: 支持多种注意力算子和量化矩阵乘法算子，系统会根据安装状态自动排序
+
+### 输入参数
+
+- **提示词 (Prompt)**: 描述期望的视频内容
+- **负向提示词 (Negative Prompt)**: 指定不希望出现的元素
+- **输入图像**: i2v 模式下需要上传输入图像
+- **分辨率**: 支持多种预设分辨率（480p/540p/720p）
+- **随机种子**: 控制生成结果的随机性
+- **推理步数**: 影响生成质量和速度的平衡（蒸馏模型默认为 4 步）
+
+### 视频参数
+
+- **FPS**: 每秒帧数
+- **总帧数**: 视频长度
+- **CFG缩放因子**: 控制提示词影响强度（1-10，蒸馏模型默认为 1）
+- **分布偏移**: 控制生成风格偏离程度（0-10）
+
+## 🔧 自动配置功能
+
+系统会根据您的硬件配置（GPU 显存和 CPU 内存）自动配置最优推理选项，无需手动调整。启动时会自动应用最佳配置，包括：
+
+- **GPU 内存优化**: 根据显存大小自动启用 CPU 卸载、VAE 分块推理等
+- **CPU 内存优化**: 根据系统内存自动启用延迟加载、模块卸载等
+- **算子选择**: 自动选择已安装的最优算子（按优先级排序）
+- **量化配置**: 根据模型文件名自动检测并应用量化精度
+
+
+### 日志查看
+
+```bash
+# 查看推理日志
+tail -f inference_logs.log
+
+# 查看GPU使用情况
+nvidia-smi
+
+# 查看系统资源
+htop
+```
+
+欢迎提交Issue和Pull Request来改进这个项目！
+
+**注意**: 使用本工具生成的视频内容请遵守相关法律法规，不得用于非法用途。