diff --git a/ci/gpu/build.sh b/ci/gpu/build.sh
index 7c9c353f23c..a74eefffc8b 100755
--- a/ci/gpu/build.sh
+++ b/ci/gpu/build.sh
@@ -72,10 +72,6 @@ gpuci_conda_retry install -y \
 # gpuci_conda_retry remove --force rapids-build-env rapids-notebook-env
 # gpuci_conda_retry install -y "your-pkg=1.0.0"
 
-gpuci_logger "Install the master version of dask and distributed"
-pip install "git+https://github.com/dask/distributed.git" --upgrade --no-deps
-pip install "git+https://github.com/dask/dask.git" --upgrade --no-deps
-
 gpuci_logger "Check versions"
 python --version
 $CC --version
@@ -108,6 +104,10 @@ else
     echo "Installing $CONDA_FILE"
     conda install -c ${CONDA_ARTIFACT_PATH} "$CONDA_FILE"
 
+    gpuci_logger "Install the master version of dask and distributed"
+    pip install "git+https://github.com/dask/distributed.git" --upgrade --no-deps
+    pip install "git+https://github.com/dask/dask.git" --upgrade --no-deps
+
     echo "Build cugraph..."
     $WORKSPACE/build.sh cugraph
 fi
diff --git a/cpp/src/community/triangles_counting.cu b/cpp/src/community/triangles_counting.cu
index 31acb4859bd..4d6f8029a56 100644
--- a/cpp/src/community/triangles_counting.cu
+++ b/cpp/src/community/triangles_counting.cu
@@ -827,7 +827,8 @@ void TrianglesCount<IndexType>::count()
     tcount_wrp();
   else {
     const int shMinBlkXSM = 6;
-    if (size_t{m_shared_mem_per_block * 8 / shMinBlkXSM} < (size_t)m_mat.N)
+    if (static_cast<size_t>(m_shared_mem_per_block * 8 / shMinBlkXSM) <
+        static_cast<size_t>(m_mat.N))
       tcount_b2b();
     else
       tcount_bsh();
diff --git a/cpp/src/components/weakly_connected_components.cu b/cpp/src/components/weakly_connected_components.cu
index 0c552ad24fc..241b7182fda 100644
--- a/cpp/src/components/weakly_connected_components.cu
+++ b/cpp/src/components/weakly_connected_components.cu
@@ -632,7 +632,7 @@ void weakly_connected_components_impl(raft::handle_t const &handle,
                                                edge_first,
                                                edge_first + new_num_edge_inserts);
         auto num_unique_edges = static_cast<size_t>(thrust::distance(edge_first, unique_edge_last));
-        num_edge_inserts.set_value(num_unique_edges, handle.get_stream_view());
+        num_edge_inserts.set_value_async(num_unique_edges, handle.get_stream_view());
       }
 
       vertex_frontier.get_bucket(static_cast<size_t>(Bucket::cur)).clear();
diff --git a/cpp/src/traversal/tsp.cu b/cpp/src/traversal/tsp.cu
index a28ddbbaa3f..17183d48f3d 100644
--- a/cpp/src/traversal/tsp.cu
+++ b/cpp/src/traversal/tsp.cu
@@ -19,6 +19,8 @@
 
 #include <utilities/high_res_timer.hpp>
 
+#include <rmm/cuda_stream_view.hpp>
+
 #include "tsp.hpp"
 #include "tsp_solver.hpp"
 
@@ -53,6 +55,8 @@ TSP::TSP(raft::handle_t const &handle,
     warp_size_(handle_.get_device_properties().warpSize),
     sm_count_(handle_.get_device_properties().multiProcessorCount),
     restart_batch_(8192),
+    mylock_scalar_(stream_),
+    best_cost_scalar_(stream_),
     neighbors_vec_((k_ + 1) * nodes_, stream_),
     work_vec_(restart_batch_ * ((4 * nodes_ + 3 + warp_size_ - 1) / warp_size_ * warp_size_),
               stream_),
@@ -81,9 +85,9 @@ void TSP::setup()
 
 void TSP::reset_batch()
 {
-  mylock_scalar_.set_value_zero(stream_);
+  mylock_scalar_.set_value_to_zero_async(stream_);
   auto const max{std::numeric_limits<int>::max()};
-  best_cost_scalar_.set_value(max, stream_);
+  best_cost_scalar_.set_value_async(max, stream_);
 }
 
 void TSP::get_initial_solution(int const batch)