Support NVFP4 quantized dense models on AMD CDNA2/CDNA3 GPUs (#7302)

Co-authored-by: HAI <hixiao@gmail.com>
Co-authored-by: Sai Enduri <saimanas.enduri@amd.com>

python/sglang/srt/layers/quantization/__init__.py

@@ -58,6 +58,7 @@ from sglang.srt.layers.quantization.modelopt_quant import (
     ModelOptFp8Config,
 )
 from sglang.srt.layers.quantization.moe_wna16 import MoeWNA16Config
+from sglang.srt.layers.quantization.petit import PetitNvFp4Config
 from sglang.srt.layers.quantization.qoq import QoQConfig
 from sglang.srt.layers.quantization.utils import get_linear_quant_method
 from sglang.srt.layers.quantization.w4afp8 import W4AFp8Config
@@ -76,6 +77,7 @@ BASE_QUANTIZATION_METHODS: Dict[str, Type[QuantizationConfig]] = {
     "compressed-tensors": CompressedTensorsConfig,
     "qoq": QoQConfig,
     "w4afp8": W4AFp8Config,
+    "petit_nvfp4": PetitNvFp4Config,
 }
 
 # VLLM-dependent quantization methods