#
Starred repositories
Diverse Score Distillation
[Preprint] Efficient Generative Model Training via Embedded Representation Warmup
Score identity Distillation with Long and Short Guidance for One-Step Text-to-Image Generation
[CVPR 2025] Attention Distillation: A Unified Approach to Visual Characteristics Transfer
[CVPR 2025 Oral] Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models
🚀🚀 「大模型」2小时完全从0训练26M的小参数GPT！🌏 Train a 26M-parameter GPT from scratch in just 2h!
The offical implementation of [NeurIPS2024] Wasserstein Distance Rivals Kullback-Leibler Divergence for Knowledge Distillation https://arxiv.org/abs/2412.08139
[NeurIPS'23 Oral] Visual Instruction Tuning (LLaVA) built towards GPT-4V level capabilities and beyond.
[IEEE T-PAMI 2024] All you need for End-to-end Autonomous Driving
A PyTorch library for implementing flow matching algorithms, featuring continuous and discrete flow matching implementations. It includes practical examples for both text and image modalities.
[NeurIPS 24] MoE Jetpack: From Dense Checkpoints to Adaptive Mixture of Experts for Vision Tasks
The codebase of our paper "Improving the Training of Rectified Flows", NeurIPS 2024
"Effective Whole-body Pose Estimation with Two-stages Distillation" (ICCV 2023, CV4Metaverse Workshop)
SpeeD: A Closer Look at Time Steps is Worthy of Triple Speed-Up for Diffusion Model Training
EDM2 and Autoguidance -- Official PyTorch implementation
High-resolution models for human tasks.
欢迎来到 LLM-Dojo，这里是一个开源大模型学习场所，使用简洁且易阅读的代码构建模型训练框架(支持各种主流模型如Qwen、Llama、GLM等等)、RLHF框架(DPO/CPO/KTO/PPO)等各种功能。👩🎓👨🎓
[ACM MM24] Official implementation of ACM MM 2024 paper: "ZePo: Zero-Shot Portrait Stylization with Faster Sampling"
Efficient distillation of diffusion-based text-to-image models using only a teacher model and prompts (no training images).
Papers and implementations of distillation for diffusion models
We are apply the distillation process for the hugging face diffusers
PyTorch code and model checkpoints for Score identity Distillation (SiD) and its adversarial version (SiDA)
(NeurIPS 2024 Oral 🔥) Improved Distribution Matching Distillation for Fast Image Synthesis
Open source implementation and models of One-step Diffusion with Distribution Matching Distillation
StyleShot: A SnapShot on Any Style. 一款可以迁移任意风格到任意内容的模型，无需针对图片微调，即能生成高质量的个性风格化图片!
[ICCV 2023] Zero-shot image editing with stochastic diffusion models
Implementation of the 2023 CVPR Award Candidate: On Distillation of Guided Diffusion Models
A universal Stable-Diffusion toolbox