Update on "[ET-VK][ez] Fix handling of assert ops"

## Changes

* Apply `RemoveAssertsTransform` as part of `vulkan_preprocess`
* Do not call `RemoveAssertsTransform` before lowering the graph
* Register ops related to asserts to the operator registry as ephemeral ops

## Motivation

assert ops are not implemented in Vulkan, so previously `RemoveAssertsTransform()` is called on the graph before the lowering process.

However, it turns out that the assertion ops are required to properly handle dynamic shapes, because they place constraints on the possible range of symbolic integers. If they are not present, then re-tracing the graph during a recompile (which may occur during a graph transform pass) may fail.

Therefore, instead of calling the transform before lowering, call it inside vulkan_preprocess after a point where subsequent passes will not attempt to trace the graph.

Differential Revision: [D75686048](https://our.internmc.facebook.com/intern/diff/D75686048/)

[ghstack-poisoned]

Author: SS-JIA

.github/scripts/label_utils.py

@@ -22,9 +22,7 @@
 
 LABEL_ERR_MSG_TITLE = "This PR needs a `release notes:` label"
 LABEL_ERR_MSG = f"""# {LABEL_ERR_MSG_TITLE}
-If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with `release notes:`.
-
-If not, please add the `release notes: none` label.
+If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with `release notes:`. This helps us keep track and include your important work in the next release notes.
 
 To add a label, you can comment to pytorchbot, for example
 `@pytorchbot label "release notes: none"`

.github/scripts/trymerge.py

@@ -59,12 +59,7 @@
     patterns_to_regex,
     retries_decorator,
 )
-from label_utils import (
-    gh_add_labels,
-    gh_remove_label,
-    has_required_labels,
-    LABEL_ERR_MSG,
-)
+from label_utils import gh_add_labels, gh_remove_label
 from trymerge_explainer import get_revert_message, TryMergeExplainer
 
 # labels
@@ -2116,9 +2111,6 @@ def merge(
     # Check for approvals
     find_matching_merge_rule(pr, repo, skip_mandatory_checks=True)
 
-    if not has_required_labels(pr):
-        raise RuntimeError(LABEL_ERR_MSG.lstrip(" #"))
-
     if ignore_current:
         checks = pr.get_checkrun_conclusions()
         _, failing, _ = categorize_checks(

.github/workflows/check-labels.yml

@@ -51,4 +51,4 @@ jobs:
           PR_NUM: ${{ github.event.number || github.event.inputs.pr_number }}
         run: |
           set -ex
-          python3 .github/scripts/check_labels.py --exit-non-zero "${PR_NUM}"
+          python3 .github/scripts/check_labels.py "${PR_NUM}"

.github/workflows/trunk.yml

@@ -305,7 +305,7 @@ jobs:
         # Install requirements
         ${CONDA_RUN} sh install_requirements.sh
         ${CONDA_RUN} sh backends/apple/coreml/scripts/install_requirements.sh
-        ${CONDA_RUN} python install_executorch.py --pybind coreml
+        ${CONDA_RUN} python install_executorch.py
         ${CONDA_RUN} sh examples/models/llama/install_requirements.sh
 
         # Test ANE llama

CMakePresets.json

@@ -15,7 +15,7 @@
         "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/third-party/ios-cmake/ios.toolchain.cmake",
         "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/macos.cmake",
         "PLATFORM": "MAC_ARM64",
-        "DEPLOYMENT_TARGET": "10.15"
+        "DEPLOYMENT_TARGET": "12.0"
       },
       "condition": {
         "lhs": "${hostSystemName}",
@@ -77,7 +77,7 @@
       "inherits": ["common"],
       "cacheVariables": {
         "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/pybind.cmake",
-        "CMAKE_OSX_DEPLOYMENT_TARGET": "10.15"
+        "CMAKE_OSX_DEPLOYMENT_TARGET": "12.0"
       },
       "condition": {
         "type": "inList",
@@ -93,7 +93,7 @@
         ],
         "cacheVariables": {
             "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/llm.cmake",
-            "CMAKE_OSX_DEPLOYMENT_TARGET": "10.15"
+            "CMAKE_OSX_DEPLOYMENT_TARGET": "12.0"
         },
         "condition": {
             "type": "inList",

Package.swift

@@ -120,7 +120,7 @@ let package = Package(
   name: "executorch",
   platforms: [
     .iOS(.v17),
-    .macOS(.v10_15),
+    .macOS(.v12),
   ],
   products: packageProducts,
   targets: packageTargets + [

backends/apple/coreml/runtime/delegate/ETCoreMLDefaultModelExecutor.mm

@@ -27,7 +27,7 @@ - (instancetype)initWithModel:(ETCoreMLModel *)model {
                                                  eventLogger:(const executorchcoreml::ModelEventLogger* _Nullable __unused)eventLogger
                                                        error:(NSError * __autoreleasing *)error {
     if (self.ignoreOutputBackings) {
-        if (@available(macOS 11.0, iOS 16.0, tvOS 16.0, watchOS 9.0, *)) {
+        if (@available(iOS 16.0, tvOS 16.0, watchOS 9.0, *)) {
             predictionOptions.outputBackings = @{};
         }
     }

backends/apple/coreml/runtime/delegate/ETCoreMLModelManager.mm

@@ -92,7 +92,7 @@ BOOL is_backed_by_same_buffer(MLMultiArray *array1, MLMultiArray *array2) {
                                             NSOrderedSet<NSString *> *output_names,
                                             NSError * __autoreleasing *error) {
     MLPredictionOptions *options = [MLPredictionOptions new];
-    if (@available(macOS 11.0, iOS 16.0, tvOS 16.0, watchOS 9.0, *)) {
+    if (@available(iOS 16.0, tvOS 16.0, watchOS 9.0, *)) {
         NSMutableDictionary<NSString *, id> *output_backings = [NSMutableDictionary dictionary];
         NSEnumerator<NSString *> *enumerator = [output_names objectEnumerator];
         for (MLMultiArray *output in outputs) {
@@ -687,7 +687,7 @@ - (void)addPrewarmedAsset:(ETCoreMLAsset *)asset {
                                                                  eventLogger:eventLogger
                                                                        error:&localError];
     // Try without output backings.
-    if (@available(macOS 11.0, iOS 16.0, tvOS 16.0, watchOS 9.0, *)) {
+    if (@available(iOS 16.0, tvOS 16.0, watchOS 9.0, *)) {
         if (!modelOutputs && predictionOptions.outputBackings.count > 0) {
             executor.ignoreOutputBackings = YES;
             localError = nil;

backends/apple/mps/setup.md

@@ -91,7 +91,7 @@ I 00:00:00.122615 executorch:mps_executor_runner.mm:501] Model verified successf
 ### [Optional] Run the generated model directly using pybind
 1. Make sure `pybind` MPS support was installed:
 ```bash
-./install_executorch.sh --pybind mps
+CMAKE_ARGS="-DEXECUTORCH_BUILD_MPS=ON" ./install_executorch.sh
 ```
 2. Run the `mps_example` script to trace the model and run it directly from python:
 ```bash

backends/arm/_passes/__init__.py

@@ -23,6 +23,7 @@
 from .decompose_cosine_similarity_pass import DecomposeCosineSimilarityPass  # noqa
 from .decompose_div_pass import DecomposeDivPass  # noqa
 from .decompose_gelu_pass import DecomposeGeluPass  # noqa
+from .decompose_groupnorm_pass import DecomposeGroupNormPass  # noqa
 from .decompose_layernorm_pass import DecomposeLayerNormPass  # noqa
 from .decompose_leaky_relu_pass import DecomposeLeakyReLUPass  # noqa
 from .decompose_linalg_vector_norm_pass import DecomposeLinearVectorNormPass  # noqa