From 2d3c0a15bb2f23d91f83bec2a896f61fc77273cf Mon Sep 17 00:00:00 2001 From: xuzhang Date: Wed, 19 Nov 2025 17:20:20 +0800 Subject: [PATCH] Update README Updated the README.md to reflect new features and releases in the PPDiffusers toolkit, including version updates and new algorithms. --- ppdiffusers/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/ppdiffusers/README.md b/ppdiffusers/README.md index b3ff4a856..bddd70f68 100644 --- a/ppdiffusers/README.md +++ b/ppdiffusers/README.md @@ -21,7 +21,7 @@ ## News 📢 * 🔥 **2025.07.14 发布 0.30.0 版本，新增文生图算法FLUX，文生视频算法[HunyuanVideo](examples/HunyuanVideo/)和[Wan2.1](examples/Wan2.1/)。发布推理加速工具包[Fast-Diffusers](examples/Fast-Diffusers)，包含SOTA Training-Free算法：[T-gate](examples/Fast-Diffusers/Training-Free/tgate), [PAB](examples/Fast-Diffusers/Training-Free/pab), [TeaCache](examples/Fast-Diffusers/Training-Free/teacache), [TaylorSeer](examples/Fast-Diffusers/Training-Free/taylorseer), [BlockDance](examples/Fast-Diffusers/Training-Free/blockdance)等和自研算法：[SortBlock](examples/Fast-Diffusers/Training-Free/sortblock)，[TeaBlockCache](examples/Fast-Diffusers/Training-Free/teablockcache), [CG-Taylor](ppdiffusers/examples/Fast-Diffusers/Training-Free/CG-Taylor/)和[FirstBlockTaylor](examples/Fast-Diffusers/Training-Free/firstblock_taylorseer)算法，在保证生成图像质量的同时，实现2倍以上的端到端推理加速效果。新增[PCM](examples/Fast-Diffusers/diffusion-distill/phased_consistency_distillation),[DMD2](examples/Fast-Diffusers/diffusion-distill//dmd2)等扩散模型蒸馏算法，并提供了多种蒸馏loss供开发者灵活搭配。同时基于上述蒸馏算法，发布了基于FLUX-dev的4步蒸馏模型，配合飞桨深度学习编译器，推理时延降低至1.66秒。** -* + * 🔥 **2025.03.11 发布 0.29.1 版本，发布自研可控视频模型 [PP-VCtrl](https://github.com/westfish/PaddleMIX/tree/develop/ppdiffusers/examples/ppvctrl)，支持在多种控制条件下的视频生成，广泛适用于人物动画、场景转换和视频编辑场景；集成高质量文生视频模型 [CogVideoX](https://github.com/westfish/PaddleMIX/tree/develop/ppdiffusers/examples/cogvideo) 和 HunyuanVideo，进一步提高了文本到视频生成的效果；新增对 SD3 ControlNet 和 SD3.5 的支持，提升了Stable Diffusion 3的兼容性，进一步强化图像生成的精细化控制能力。** * 🔥 **2024.10.18 发布 0.29.0 版本，新增图像生成模型[Stable Diffusion 3 (SD3)](https://github.com/PaddlePaddle/PaddleMIX/blob/develop/ppdiffusers/examples/text_to_image/README_sd3.md)，支持DreamBooth训练及高性能推理；SD3、SDXL适配昇腾910B，提供国产计算芯片上的训推能力；DIT支持[高性能推理](https://github.com/PaddlePaddle/PaddleMIX/blob/develop/ppdiffusers/examples/class_conditional_image_generation/DiT/README.md#23-paddle-inference-%E9%AB%98%E6%80%A7%E8%83%BD%E6%8E%A8%E7%90%86)；支持PaddleNLP 3.0 beta版本。**