[AMD][MXFP4] Online MXFP4 quantization 1/N - dense and MOE models w. original BF16 weight (#18005)

Co-authored-by: Bowen Bao <bowenbao@amd.com>
Co-authored-by: Colin Zeng <Colin.Zeng@amd.com>

docs_new/docs/advanced_features/quantization.mdx

@@ -14,7 +14,7 @@ on-the-fly to convert high-precision weights into a lower-precision format.
 
 **Note: For better performance, usability and convenience, offline quantization is recommended over online quantization.**
 
-If you use a pre-quantized model, do not add `--quantization` to enable online quantization at the same time.
+If you use a pre-quantized model, **do not add `--quantization` to enable online quantization at the same time**.
 For popular pre-quantized models, please visit [Unsloth](https://huggingface.co/unsloth), [NVIDIA ModelOpt](https://huggingface.co/collections/nvidia/inference-optimized-checkpoints-with-model-optimizer)
 or [NeuralMagic](https://huggingface.co/collections/neuralmagic) collections on HF for some
 popular quality validated quantized models. Quantized models must be validated via benchmarks post-quantization
@@ -804,6 +804,18 @@ SGLang running on AMD GPUs (CDNA3 or CDNA4 architecture) supports the quantizati
 
 Other layers (e.g. projections in the attention layers) have their weights quantized online to float8 directly.
 
+### `quark_mxfp4` online quantization method
+
+SGLang running on AMD GPUs with hardware FP4 support (CDNA4 architecture, e.g. MI355x) supports the quantization method `--quantization quark_mxfp4`, that will quantize BF16 model weights to MXFP4 at load time, use dynamic MXFP4 quantization for activations and MXFP4 GEMMs instead of BF16 GEMMs.
+
+Example:
+
+```bash
+sglang serve --model-path Qwen/Qwen3-30B-A3B \
+    --tensor-parallel-size 1 \
+    --quantization quark_mxfp4
+```
+
 ## Reference
 
 - [GPTQModel](https://github.com/ModelCloud/GPTQModel)