Enhance README with vLLM/SGLang examples

README.md

@@ -338,7 +338,7 @@ Install sglang-diffusion from source:
 pip install git+https://github.com/sgl-project/sglang.git
 ```
 
-For example, let's take Helios-Base.
+For example, let's take Helios-Base. **(Native Support)**
 
 <details>
   <summary>Click to expand the code</summary>
@@ -356,6 +356,25 @@ For example, let's take Helios-Base.
   ```
 </details>
 
+For example, let's take Helios-Base. **(Diffusers Backend)**
+
+<details>
+  <summary>Click to expand the code</summary>
+
+  ```bash
+  sglang generate \
+    --model-path BestWishYsh/Helios-Base \
+    --prompt "A cat walking on the beach at sunset, cinematic lighting, high quality" \
+    --negative-prompt "Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards" \
+    --height 384 \
+    --width 640 \
+    --num-frames 33 \
+    --num-inference-steps 50 \
+    --guidance-scale 5.0 \
+    ----backend diffusers
+  ```
+</details>
+
 ## 🗝️ Training
 
 We use a three-stage progressive pipeline, all the setting can be found [here](./scripts/training/configs/). Stage-1 (Base) performs architectural adaptation: we apply Unified History Injection, Easy Anti-Drifting, and Multi-Term Memory Patchification to convert the bidirectional pretrained model into an autoregressive generator. Stage-2 (Mid) targets token compression by introducing Pyramid Unified Predictor Corrector, which aggressively reduces the number of noisy tokens and thus the overall computation. Stage-3 (Distilled) applies Adversarial Hierarchical Distillation, reducing the sampling steps from 50 to 3 and eliminating the need for classifier-free guidance (CFG). Throughout training, we apply dynamic shifting to all timestep-dependent operations to match the noise schedule to the latent size.