README: scope the download, fix the single-GPU vLLM-Omni command

README.md

@@ -127,7 +127,7 @@ into the image.
 | `pdd_fused_heads.safetensors`, `pdd_config.json` | The per-step output heads that replace the transformer's output layer, and the sampling schedule. | 12 MB |
 | `vae/` | Image VAE, FP32. | 2.34 GB |
 | `tokenizer/`, `processor/` | Qwen3-VL tokenizer and processor, identical to MiniMax-H3. | 23 MB |
-| Text encoder | Not in this repository. `model_index.json` loads it from [`MiniMaxAI/MiniMax-H3`](https://huggingface.co/MiniMaxAI/MiniMax-H3), folder `text_encoder/`. If you already have MiniMax-H3, it is not downloaded again. | 66.7 GB |
+| Text encoder | Not part of the original checkpoint. `model_index.json` loads it from [`MiniMaxAI/MiniMax-H3`](https://huggingface.co/MiniMaxAI/MiniMax-H3), folder `text_encoder/`. If you already have MiniMax-H3, it is not downloaded again. | 66.7 GB |
 
 ```text
 MiniMax-H3-Image/
@@ -140,15 +140,21 @@ MiniMax-H3-Image/
 └── vae/
 ```
 
-Download this repository and the MiniMax-H3 text encoder into the Hugging Face cache:
+The repository hosts this original checkpoint and the diffusers format side by side, so scope the
+download to what your framework needs. Download the original checkpoint (SGLang, vLLM-Omni) and
+the MiniMax-H3 text encoder into the Hugging Face cache:
 
 ```bash
-hf download MiniMaxAI/MiniMax-H3-Image
+hf download MiniMaxAI/MiniMax-H3-Image model_index.json "pdd_*" "processor/*" "tokenizer/*" "transformer/*" "vae/*"
 hf download MiniMaxAI/MiniMax-H3 --include "text_encoder/*"
 # optional prompt expander
 hf download MiniMaxAI/MiniMax-H3-Image-PE
 ```
 
+diffusers users do not need a manual download:
+`ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3-Image")` fetches the diffusers-format
+folders it needs.
+
 ### Local Deployment
 
 We recommend the following inference frameworks to serve the model:
@@ -164,12 +170,19 @@ See the [vLLM-Omni recipe](https://recipes.vllm.ai/MiniMaxAI/MiniMax-H3-Image) f
 per-hardware deployment configurations and measured latency. Both NVIDIA and AMD accelerators are
 supported, including NVIDIA H200, RTX PRO 6000 Blackwell, AMD Instinct MI350X and more.
 
-Start the server on one GPU:
+Start the server on one GPU with about 135 GiB of free memory, such as B200 or B300:
 
 ```bash
 vllm serve MiniMaxAI/MiniMax-H3-Image --omni --host 0.0.0.0 --port 8000
 ```
 
+On one H200, offload the text encoder:
+
+```bash
+vllm serve MiniMaxAI/MiniMax-H3-Image --omni --host 0.0.0.0 --port 8000 --num-gpus 1 \
+  --diffusion-offload-config '{"mode":"layer","components":["text_encoder"]}'
+```
+
 For lower latency, split each request over several GPUs with sequence parallelism
 (Ulysses), for example on 4 GPUs: