Skip to content

Arm backend: Upgrade model converter to 0.10.0 #21809

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wwwind wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Arm backend: Upgrade model converter to 0.10.0 #21809

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cc31afb
Arm backend: Upgrade model converter to 0.10.0
wwwind Aug 7, 2026
196a8dd
Merge branch 'main' into vgf_upgrade
wwwind Aug 13, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion backends/arm/requirements-arm-vgf-runtime.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,4 +5,4 @@

# Runtime build dependencies for the Arm VGF backend.

ai_ml_sdk_vgf_library == 0.9.0
ai_ml_sdk_vgf_library == 0.10.0
4 changes: 2 additions & 2 deletions backends/arm/requirements-arm-vgf.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,5 +5,5 @@

# These dependencies need to match pyproject.toml

ai_ml_emulation_layer_for_vulkan == 0.9.0
ai_ml_sdk_model_converter == 0.9.0
ai_ml_emulation_layer_for_vulkan == 0.10.0
ai_ml_sdk_model_converter == 0.10.0
30 changes: 6 additions & 24 deletions backends/arm/runtime/VGFBackend.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -100,21 +100,6 @@ VkResult vkml_allocate_basics(
VkCommandPool* command_pool,
uint32_t* queue_family_index);

void vkml_free_basics(
VkInstance* instance,
VkDevice* device,
VkCommandPool* command_pool) {
if (*device != VK_NULL_HANDLE && *command_pool != VK_NULL_HANDLE) {
vkDestroyCommandPool(*device, *command_pool, nullptr);
}
// Note: These primitives are used by the emulation layer for vulkan
// object allocation, the vulkan objects are freed in in library
// shutdown, so we can't yet destroy these here without causing
// a crash there.
// vkDestroyDevice(*device, nullptr);
// vkDestroyInstance(*instance, nullptr);
}

// Helper functions to dump VGF Delegate Boundary Inputs
constexpr const char* kVgfDumpInputsDirEnv = "EXECUTORCH_VGF_DUMP_INPUTS_DIR";
constexpr const char* kVgfDumpInputsAndExitEnv =
Expand Down Expand Up @@ -456,9 +441,8 @@ class VGFBackend final : public ::executorch::runtime::BackendInterface {

is_initialized_ = true;
}
~VGFBackend() {
vkml_free_basics(&vk_instance, &vk_device, &vk_command_pool);
}

~VGFBackend() = default;

bool is_available() const override {
ET_LOG(Info, "Checking VGFBackend is available");
Expand Down Expand Up @@ -621,12 +605,10 @@ class VGFBackend final : public ::executorch::runtime::BackendInterface {
++input_arg_idx) {
const int io_idx = repr->model_input_io_index[input_arg_idx];
if (io_idx < 0) {
#ifdef ET_EVENT_TRACER_ENABLED
event_tracer_end_profiling_delegate(event_tracer, copy_inputs_event);
event_tracer_end_profiling_delegate(event_tracer, vgf_execute_event);
#endif
ET_LOG(Error, "Missing IO mapping for input %zu", input_arg_idx);
return Error::InvalidArgument;
ET_LOG(Info, "Skipping eliminated VGF input %zu", input_arg_idx);
// See test_addmm_vgf_no_quant[beta_only]
// two inputs are eliminated from the graph by the converter
continue;
}
if (!args[input_arg_idx]->isTensor()) {
#ifdef ET_EVENT_TRACER_ENABLED
Expand Down
Loading
Loading