diff --git a/docs_new/cards/logos/ideogram.png b/docs_new/cards/logos/ideogram.png index f39ef1145..0742f783f 100644 Binary files a/docs_new/cards/logos/ideogram.png and b/docs_new/cards/logos/ideogram.png differ diff --git a/docs_new/cookbook/diffusion/LingBot-World/LingBot-World.mdx b/docs_new/cookbook/diffusion/LingBot-World/LingBot-World.mdx index b72b284f9..16b8c6cfa 100644 --- a/docs_new/cookbook/diffusion/LingBot-World/LingBot-World.mdx +++ b/docs_new/cookbook/diffusion/LingBot-World/LingBot-World.mdx @@ -5,36 +5,16 @@ metatags: tag: REALTIME --- +
+ realtime + world model + causal DiT +
+ ## 1. Model Introduction [LingBot World](https://huggingface.co/robbyant/lingbot-world-fast-diffusers) is a realtime camera-controlled video world model. In SGLang-diffusion, it belongs to the realtime causal path: the server keeps a live session, samples control signals per chunk, reuses causal DiT state, and decodes video frames incrementally. -
-
-
-
LingBot World
-
Realtime diffusion world model in SGLang-diffusion
-
-
- prompt + image + control → streaming video -
-
-
-
Category
-
- realtime - world model - causal DiT -
-
Inputs
-
Prompt, first frame, and per-chunk camera control signals
-
Outputs
-
Streaming video frame chunks over /v1/realtime_video/generate
-
Core runtime
-
Condition queue, causal DiT KV cache, causal VAE decode cache, and realtime session state
-
-
- This is different from offline diffusion video models such as Wan or LTX. Offline models denoise a bounded latent sequence for one request. Realtime world models generate a continuing stream, so the runtime must manage session state, control events, causal attention cache, and VAE decode cache. ## 2. Deployment diff --git a/docs_new/cookbook/diffusion/Wan/Wan2.1.mdx b/docs_new/cookbook/diffusion/Wan/Wan2.1.mdx index 7cafe28fc..479ff5b95 100644 --- a/docs_new/cookbook/diffusion/Wan/Wan2.1.mdx +++ b/docs_new/cookbook/diffusion/Wan/Wan2.1.mdx @@ -35,7 +35,7 @@ This section provides deployment configurations optimized for different hardware ### 3.1 Basic Configuration -The Wan2.1 series offers models in multiple sizes and resolutions, optimized for different hardware platforms. The recommended launch configurations vary by hardware and model size. +The Wan2.1 series offers models in multiple sizes and resolutions. SGLang supports Wan2.1 deployment on NVIDIA B200, B300, H200, H100, and AMD MI300X, MI325X, MI355X GPUs. The recommended launch configurations vary by hardware, model size, and memory headroom. **Interactive Command Generator**: Use the configuration selector below to automatically generate an appropriate deployment command for your model variant and options. @@ -169,7 +169,9 @@ sglang serve --model-path Wan-AI/Wan2.1-T2V-14B-Diffusers --port 30000 \ --lora-path NIVEDAN/wan2.1-lora ``` -## 5. Benchmark +## 5. Reference Benchmark + +The following benchmark is a point-in-time reference for one model, hardware platform, SGLang image, and parameter set. It is not a complete hardware support matrix. Test Environment: diff --git a/docs_new/cookbook/diffusion/intro.mdx b/docs_new/cookbook/diffusion/intro.mdx index 35dbf76a8..a09975fad 100644 --- a/docs_new/cookbook/diffusion/intro.mdx +++ b/docs_new/cookbook/diffusion/intro.mdx @@ -6,17 +6,13 @@ metatags: description: "Explore SGLang diffusion model cookbooks for image and video generation deployment, invocation, optimization, and benchmarking examples." --- -## Offline Diffusion Models +Choose a recipe by output modality. The sidebar stays organized by model family, while this overview separates image, video, and realtime/world workloads. -Offline models generate each image or video request as a bounded denoising job. Most of them use bidirectional attention over the whole latent sequence. +## Image Models + +Image models generate one image request as a bounded denoising job, usually with bidirectional attention over the whole latent sequence. - - - + + +## Video Models + +Video models denoise a bounded latent video sequence for each request. Use these recipes for offline text-to-video, image-to-video, and video generation serving. + + + + + + { hardware: { name: 'hardware', title: 'Hardware Platform', - items: [{ id: 'mi300x', label: 'MI300X/MI325X/MI355X', default: true }], + items: [ + { id: 'b200', label: 'B200', default: true }, + { id: 'b300', label: 'B300', default: false }, + { id: 'h200', label: 'H200', default: false }, + { id: 'h100', label: 'H100', default: false }, + { id: 'mi300x', label: 'MI300X', default: false }, + { id: 'mi325x', label: 'MI325X', default: false }, + { id: 'mi355x', label: 'MI355X', default: false }, + ], }, task: { name: 'task', @@ -77,7 +85,7 @@ export const Wan21Deployment = () => { const configKey = `${task}-${modelsize}`; const supported = modelConfigs[configKey]?.supportedLoras || []; return { - hardware: 'mi300x', + hardware: 'b200', task, modelsize, bestPractice: 'off',