diff --git a/backends/vulkan/runtime/api/Context.cpp b/backends/vulkan/runtime/api/Context.cpp
index 326391424df..80aef97fc04 100644
--- a/backends/vulkan/runtime/api/Context.cpp
+++ b/backends/vulkan/runtime/api/Context.cpp
@@ -226,13 +226,14 @@ void Context::submit_cmd_to_gpu(VkFence fence_handle, const bool final_use) {
   }
 }
 
-void Context::flush() {
+void Context::wait_for_queue() {
   VK_CHECK(vkQueueWaitIdle(queue().handle));
+}
 
+void Context::clear_resources() {
   command_pool_.flush();
   descriptor_pool_.flush();
 
-  // If there is an existing command buffer, invalidate it
   if (cmd_) {
     cmd_.invalidate();
   }
@@ -243,6 +244,11 @@ void Context::flush() {
   images_to_clear_.clear();
 }
 
+void Context::flush() {
+  wait_for_queue();
+  clear_resources();
+}
+
 bool available() {
   return context();
 }
diff --git a/backends/vulkan/runtime/api/Context.h b/backends/vulkan/runtime/api/Context.h
index 5764cb6a894..49003435e3b 100644
--- a/backends/vulkan/runtime/api/Context.h
+++ b/backends/vulkan/runtime/api/Context.h
@@ -232,6 +232,10 @@ class Context final {
     return cmd_;
   }
 
+  void wait_for_queue();
+
+  void clear_resources();
+
   void flush();
 
 #if defined(VK_KHR_pipeline_executable_properties) && \
diff --git a/backends/vulkan/runtime/graph/ComputeGraph.cpp b/backends/vulkan/runtime/graph/ComputeGraph.cpp
index 3a5aabb1c7d..bb2df30a174 100644
--- a/backends/vulkan/runtime/graph/ComputeGraph.cpp
+++ b/backends/vulkan/runtime/graph/ComputeGraph.cpp
@@ -183,13 +183,22 @@ ComputeGraph::ComputeGraph(GraphConfig config)
 }
 
 ComputeGraph::~ComputeGraph() {
-  values_.clear();
+  // Wait for all currently executing commands to complete before cleaning up.
+  // If wait_for_queue() throws an exception, still proceed with cleanup.
+  try {
+    context_->wait_for_queue();
+  } catch (...) {
+  }
 
-  prepack_nodes_.clear();
-  execute_nodes_.clear();
-  clear_deferred_cmds();
+  // Wrap in try/catch to ensure that destructor does not throw
+  try {
+    values_.clear();
 
-  context_->flush();
+    prepack_nodes_.clear();
+    execute_nodes_.clear();
+    clear_deferred_cmds();
+  } catch (...) {
+  }
 }
 
 std::vector<int64_t> ComputeGraph::extract_int_or_symint_list(
diff --git a/backends/vulkan/runtime/vk_api/Pipeline.cpp b/backends/vulkan/runtime/vk_api/Pipeline.cpp
index 6fa85924223..522c4b8589b 100644
--- a/backends/vulkan/runtime/vk_api/Pipeline.cpp
+++ b/backends/vulkan/runtime/vk_api/Pipeline.cpp
@@ -459,7 +459,22 @@ void ComputePipelineCache::create_pipelines(
     const std::unordered_set<Key, Hasher>& descriptors) {
   std::lock_guard<std::mutex> lock(cache_mutex_);
 
-  const auto num_pipelines = descriptors.size();
+  // Filter out descriptors already in cache to avoid creating duplicate
+  // pipelines and to ensure correct indexing between created pipelines and
+  // cache insertion.
+  std::vector<Key> keys_to_create;
+  keys_to_create.reserve(descriptors.size());
+  for (const auto& key : descriptors) {
+    if (cache_.find(key) == cache_.cend()) {
+      keys_to_create.push_back(key);
+    }
+  }
+
+  if (keys_to_create.empty()) {
+    return;
+  }
+
+  const auto num_pipelines = keys_to_create.size();
   std::vector<VkPipeline> pipelines(num_pipelines);
 
   std::vector<std::vector<VkSpecializationMapEntry>> map_entries;
@@ -474,7 +489,7 @@ void ComputePipelineCache::create_pipelines(
   std::vector<VkComputePipelineCreateInfo> create_infos;
   create_infos.reserve(num_pipelines);
 
-  for (auto& key : descriptors) {
+  for (const auto& key : keys_to_create) {
     map_entries.push_back(key.specialization_constants.generate_map_entries());
 
     specialization_infos.push_back(VkSpecializationInfo{
@@ -513,14 +528,10 @@ void ComputePipelineCache::create_pipelines(
       nullptr,
       pipelines.data()));
 
-  uint32_t i = 0;
-  for (auto& key : descriptors) {
-    auto it = cache_.find(key);
-    if (it != cache_.cend()) {
-      continue;
-    }
-    cache_.insert({key, ComputePipelineCache::Value(device_, pipelines[i])});
-    ++i;
+  for (size_t i = 0; i < keys_to_create.size(); ++i) {
+    cache_.insert(
+        {keys_to_create[i],
+         ComputePipelineCache::Value(device_, pipelines[i])});
   }
 }