kompute: fix device name leaks

Signed-off-by: Jared Van Bortel <jared@nomic.ai>

gpt4all-backend/llama.cpp-mainline

@@ -1 +1 @@
-Subproject commit e21d0b32e9324b1df40107d9004cc6b67cdd1567
+Subproject commit 40bac11e427f2307305b86c322cb366bb95fcb8a

gpt4all-backend/llamamodel.cpp

@@ -405,7 +405,7 @@ bool LLamaModel::loadModel(const std::string &modelPath, int n_ctx, int ngl)
     if (usingGPUDevice()) {
 #ifdef GGML_USE_KOMPUTE
         if (llama_verbose()) {
-            std::cerr << "llama.cpp: using Vulkan on " << ggml_vk_current_device().name << std::endl;
+            std::cerr << "llama.cpp: using Vulkan on " << d_ptr->deviceName << std::endl;
         }
         d_ptr->backend_name = "kompute";
 #elif defined(GGML_USE_VULKAN)
@@ -652,6 +652,8 @@ bool LLamaModel::initializeGPUDevice(size_t memoryRequired, const std::string &n
     bool ok = ggml_vk_get_device(&device, memoryRequired, name.c_str());
     if (ok) {
         d_ptr->device = device.index;
+        d_ptr->deviceName = device.name;
+        ggml_vk_device_destroy(&device);
         return true;
     }
 #else
@@ -665,17 +667,15 @@ bool LLamaModel::initializeGPUDevice(int device, std::string *unavail_reason) co
 {
 #if defined(GGML_USE_KOMPUTE) || defined(GGML_USE_VULKAN) || defined(GGML_USE_CUDA)
     (void)unavail_reason;
-    d_ptr->device = device;
-#ifndef GGML_USE_KOMPUTE
     auto devices = availableGPUDevices();
     auto it = std::find_if(devices.begin(), devices.end(), [device](auto &dev) { return dev.index == device; });
+    d_ptr->device = device;
     d_ptr->deviceName = it < devices.end() ? it->name : "(unknown)";
-#endif
     return true;
 #else
     (void)device;
     if (unavail_reason) {
-        *unavail_reason = "built without Kompute";
+        *unavail_reason = "built without a GPU backend";
     }
     return false;
 #endif
@@ -714,9 +714,7 @@ const char *LLamaModel::backendName() const {
 
 const char *LLamaModel::gpuDeviceName() const {
     if (usingGPUDevice()) {
-#ifdef GGML_USE_KOMPUTE
-        return ggml_vk_current_device().name;
-#elif defined(GGML_USE_VULKAN) || defined(GGML_USE_CUDA)
+#if defined(GGML_USE_KOMPUTE) || defined(GGML_USE_VULKAN) || defined(GGML_USE_CUDA)
         return d_ptr->deviceName.c_str();
 #endif
     }

gpt4all-backend/llmodel.h

@@ -1,6 +1,7 @@
 #ifndef LLMODEL_H
 #define LLMODEL_H
 
+#include <algorithm>
 #include <cstdint>
 #include <fstream>
 #include <functional>
@@ -8,6 +9,7 @@
 #include <optional>
 #include <string>
 #include <string_view>
+#include <unordered_map>
 #include <vector>
 
 using namespace std::string_literals;
@@ -54,8 +56,24 @@ public:
             backend(backend), index(index), type(type), heapSize(heapSize), name(std::move(name)),
             vendor(std::move(vendor)) {}
 
-        std::string selectionName() const { return backend == "cuda"s ? "CUDA: " + name  : name; }
-        std::string reportedName()  const { return backend == "cuda"s ? name + " (CUDA)" : name; }
+        std::string selectionName() const { return m_backendNames.at(backend) + ": " + name; }
+        std::string reportedName()  const { return name + " (" + m_backendNames.at(backend) + ")"; }
+
+        static std::string updateSelectionName(const std::string &name) {
+            if (name == "Auto" || name == "CPU" || name == "Metal")
+                return name;
+            auto it = std::find_if(m_backendNames.begin(), m_backendNames.end(), [&name](const auto &entry) {
+                return name.starts_with(entry.second + ": ");
+            });
+            if (it != m_backendNames.end())
+                return name;
+            return "Vulkan: " + name; // previously, there were only Vulkan devices
+        }
+
+    private:
+        static inline const std::unordered_map<std::string, std::string> m_backendNames {
+            {"cuda", "CUDA"}, {"kompute", "Vulkan"},
+        };
     };
 
     class Implementation {

gpt4all-chat/mysettings.cpp

@@ -790,7 +790,23 @@ QString MySettings::device() const
 {
     QSettings setting;
     setting.sync();
-    return setting.value("device", default_device).toString();
+    auto value = setting.value("device");
+    if (!value.isValid())
+        return default_device;
+
+    auto device = value.toString();
+    if (!device.isEmpty()) {
+        auto deviceStr = device.toStdString();
+        auto newNameStr = LLModel::GPUDevice::updateSelectionName(deviceStr);
+        if (newNameStr != deviceStr) {
+            auto newName = QString::fromStdString(newNameStr);
+            qWarning() << "updating device name:" << device << "->" << newName;
+            device = newName;
+            setting.setValue("device", device);
+            setting.sync();
+        }
+    }
+    return device;
 }
 
 void MySettings::setDevice(const QString &u)