fix(mge/dnn): fix rng and topk oom in distributed training

GitOrigin-RevId: 9841d1219e113f2d8ef8e4231248ec77a997163a

src/opr/impl/misc.cpp

@@ -379,6 +379,8 @@ void TopK::init_output_static_infer_desc() {
     }
 
     auto infer_workspace = [this](TensorShape& dst, const InpVal& iv) {
+        // active comp_node for cuda launch kernel in get_workspace_in_bytes
+        comp_node().activate();
         auto k = iv.val[3].value().ptr<int>()[0];
         auto size = megdnn_opr()->get_workspace_in_bytes(
                 k, {iv.val[0].shape(), input(0)->dtype()},

src/opr/impl/rand.cpp

@@ -60,6 +60,8 @@ cg::OperatorNodeBase::NodeProp* RNGOprBase::do_make_node_prop() const {
 
 void RNGOprBase::ensure_megdnn_opr() {
     if (!m_megdnn_opr || m_megdnn_opr.comp_node() != comp_node()) {
+        // activate comp_node for curandCreateGenerator in create_megdnn_opr
+        comp_node().activate();
         m_megdnn_opr = create_megdnn_opr();
     }
 }