[ROCm][FP8] Fix for adjustments needed only for fnuz (vllm-project#14689)

Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Signed-off-by: Louis Ulmer <ulmerlouis@gmail.com>

Author: gshtras

vllm/model_executor/layers/quantization/kv_cache.py

@@ -50,7 +50,7 @@ def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
                 # We prefer to use separate k_scale and v_scale if present
                 k_scale = layer.k_scale.to("cpu").tolist()
                 v_scale = layer.v_scale.to("cpu").tolist()
-                if current_platform.is_rocm():
+                if current_platform.is_fp8_fnuz():
                     k_scale *= 2
                     v_scale *= 2
             elif layer.k_scale < 0.0 and layer.v_scale < 0.0:
@@ -66,7 +66,7 @@ def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
                 scale_to_duplicate = max(layer.k_scale, layer.v_scale)
                 k_scale = scale_to_duplicate.to("cpu").tolist()
                 v_scale = scale_to_duplicate.to("cpu").tolist()
-                if current_platform.is_rocm():
+                if current_platform.is_fp8_fnuz():
                     k_scale *= 2
                     v_scale *= 2