[AMD] Fix weight load shape mismatch for amd dsr1 0528 mxfp4 (#19425)

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

@@ -52,6 +52,7 @@ class QuarkConfig(QuantizationConfig):
         self.kv_cache_group = kv_cache_group
         self.kv_cache_config = kv_cache_config
         self.pack_method = pack_method
+        self.exclude_layers = cast(list[str], self.quant_config.get("exclude", []))
 
         self.packed_modules_mapping = self.quant_config["packed_modules_mapping"]
 
@@ -69,13 +70,17 @@ class QuarkConfig(QuantizationConfig):
     def get_name(self) -> str:
         return "quark"
 
+    def apply_weight_name_mapper(self, hf_to_sglang_mapper):
+        self.exclude_layers = hf_to_sglang_mapper.apply_list(self.exclude_layers)
+
     def get_quant_method(
         self, layer: torch.nn.Module, prefix: str
     ) -> Optional["QuantizeMethodBase"]:
         # Check if the layer is skipped for quantization.
-        exclude_layers = cast(list[str], self.quant_config.get("exclude"))
         if should_ignore_layer(
-            prefix, ignore=exclude_layers, fused_mapping=self.packed_modules_mapping
+            prefix,
+            ignore=self.exclude_layers,
+            fused_mapping=self.packed_modules_mapping,
         ):
             if isinstance(layer, LinearBase):
                 return UnquantizedLinearMethod()

python/sglang/srt/models/deepseek_nextn.py

@@ -50,6 +50,7 @@ from sglang.srt.layers.vocab_parallel_embedding import (
 from sglang.srt.model_executor.forward_batch_info import ForwardBatch
 from sglang.srt.models.deepseek_common.utils import enable_nextn_moe_bf16_cast_to_fp8
 from sglang.srt.models.deepseek_v2 import DeepseekV2DecoderLayer, DeepseekV3ForCausalLM
+from sglang.srt.models.utils import WeightsMapper
 from sglang.srt.server_args import get_global_server_args
 from sglang.srt.utils import BumpAllocator, add_prefix, is_cuda, is_npu
 
@@ -61,6 +62,7 @@ _is_npu = is_npu()
 
 
 class DeepseekModelNextN(nn.Module):
+
     def __init__(
         self,
         config: PretrainedConfig,
@@ -192,6 +194,15 @@ class DeepseekModelNextN(nn.Module):
 
 class DeepseekV3ForCausalLMNextN(DeepseekV3ForCausalLM):
 
+    # Support amd/DeepSeek-R1-0528-MXFP4 renaming: model.layers.61*.
+    # Ref: HF config.json for amd/DeepSeek-R1-0528-MXFP4
+    # https://huggingface.co/amd/DeepSeek-R1-0528-MXFP4/blob/main/config.json
+    hf_to_sglang_mapper = WeightsMapper(
+        orig_to_new_substr={
+            "model.layers.61": "model.decoder",
+        },
+    )
+
     def __init__(
         self,
         config: PretrainedConfig,