[Bugfix] fix activation in cpu_fused_moe_torch call (#34696)

Signed-off-by: Marek Michalowski <marek.michalowski@arm.com>

vllm/model_executor/layers/fused_moe/cpu_fused_moe.py

@@ -402,7 +402,7 @@ class CPUFusedMOE:
             input,
             topk_weights,
             topk_ids,
-            activation,
+            activation.value,
             global_num_experts,
             skip_weighted,
         )