[Bugfix] GPT OSS Attritbute error on H100 (#25228)

Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>

vllm/model_executor/layers/fused_moe/config.py

 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, Optional, Union
+from typing import Optional, Union
 
 import torch
 
@@ -14,7 +14,7 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
 from vllm.utils import cdiv, has_triton_kernels
 from vllm.utils.flashinfer import has_flashinfer_cutlass_fused_moe
 
-if TYPE_CHECKING and has_triton_kernels:
+if has_triton_kernels():
     from triton_kernels.matmul_ogs import PrecisionConfig
 
 logger = init_logger(__name__)

vllm/model_executor/layers/quantization/mxfp4.py

@@ -638,8 +638,8 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
             return None
 
         if self.mxfp4_backend == Mxfp4Backend.TRITON:
-            w1_scale = layer.w13_precision_config
-            w2_scale = layer.w2_precision_config
+            w1_scale = self.w13_precision_config
+            w2_scale = self.w2_precision_config
         else:
             w1_scale = layer.w13_weight_scale
             w2_scale = layer.w2_weight_scale