Update

[ghstack-poisoned]

Author: GregoryComer

.ci/scripts/utils.sh

@@ -131,8 +131,6 @@ build_executorch_runner_cmake() {
   else
       CXXFLAGS=""
   fi
-  # This command uses buck2 to gather source files and buck2 could crash flakily
-  # on MacOS
   CXXFLAGS="$CXXFLAGS" retry cmake -DPYTHON_EXECUTABLE="${PYTHON_EXECUTABLE}" -DCMAKE_BUILD_TYPE="${1:-Release}" ..
   popd || return
 

CMakeLists.txt

@@ -79,11 +79,6 @@ if(NOT PYTHON_EXECUTABLE)
 endif()
 announce_configured_options(PYTHON_EXECUTABLE)
 
-if(NOT BUCK2)
-  resolve_buck2()
-endif()
-announce_configured_options(BUCK2)
-
 announce_configured_options(CMAKE_CXX_COMPILER_ID)
 announce_configured_options(CMAKE_TOOLCHAIN_FILE)
 announce_configured_options(BUILD_TESTING)
@@ -318,17 +313,6 @@ if(EXECUTORCH_SRCS_FILE)
   )
 endif()
 executorch_load_build_variables()
-if(NOT EXECUTORCH_SRCS_FILE)
-  # A file wasn't provided. Run a script to extract the source lists from the
-  # buck2 build system and write them to a file we can include.
-  #
-  # NOTE: This will only happen once during cmake setup, so it will not re-run
-  # if the buck2 targets change.
-  message(STATUS "executorch: Generating source lists")
-  set(EXECUTORCH_SRCS_FILE "${CMAKE_CURRENT_BINARY_DIR}/executorch_srcs.cmake")
-  extract_sources(${EXECUTORCH_SRCS_FILE})
-  executorch_validate_build_variables()
-endif()
 
 # Detect if an iOS toolchain is set.
 if(CMAKE_TOOLCHAIN_FILE MATCHES ".*(iOS|ios\.toolchain)\.cmake$")

backends/arm/quantizer/arm_quantizer.py

@@ -101,18 +101,20 @@ def get_symmetric_quantization_config(
     weight_observer_or_fake_quant_ctr: ObserverOrFakeQuantizeConstructor = (
         MinMaxObserver
     )
+
     # Determine the right observer/fake-quant constructor
     if is_qat:
-        # Set plain fake-quant with true min/max
-        weight_observer_or_fake_quant_ctr = FakeQuantize
+        if is_per_channel:
+            weight_observer_or_fake_quant_ctr = PerChannelMinMaxObserver
+        else:
+            # Set plain fake-quant with true min/max
+            weight_observer_or_fake_quant_ctr = FakeQuantize
     else:
         # PTQ: set min/max observer
         weight_observer_or_fake_quant_ctr = (
             PerChannelMinMaxObserver if is_per_channel else MinMaxObserver
         )
 
-    extra_args = {"eps": 2**-12}
-
     weight_quantization_spec = QuantizationSpec(
         dtype=torch.int8,
         quant_min=weight_qmin,

backends/arm/test/misc/test_bn_relu_folding_qat.py

@@ -40,13 +40,17 @@ def forward(self, x: torch.Tensor):
 
 
 models = {
-    "conv_bn_relu": ConvModule(batch_norm=True),
-    "conv_relu": ConvModule(batch_norm=False),
+    # name : (model, is_per_channel)
+    "conv_bn_relu_per_channel": (ConvModule(batch_norm=True), True),
+    "conv_relu_per_channel": (ConvModule(batch_norm=False), True),
+    "conv_bn_relu_per_tensor": (ConvModule(batch_norm=True), False),
+    "conv_relu_per_tensor": (ConvModule(batch_norm=False), False),
 }
 
 
-@common.parametrize("model", models)
-def test_qat_tosa_INT(model: torch.nn.Module):
+@common.parametrize("test_data", models)
+def test_qat_tosa_INT(test_data):
+    model, per_channel = test_data
     pipeline = TosaPipelineINT[input_t1](model, model.test_data, [], [], qtol=1)
     tosa_version = conftest.get_option("tosa_version")
     tosa_profiles = {
@@ -59,7 +63,7 @@ def test_qat_tosa_INT(model: torch.nn.Module):
         Quantize(
             quantizer=quantizer,
             quantization_config=get_symmetric_quantization_config(
-                is_qat=True, is_per_channel=False
+                is_qat=True, is_per_channel=per_channel
             ),
             is_qat=True,
         ),

backends/cadence/build_cadence_fusionG3.sh

@@ -36,7 +36,7 @@ if $STEPWISE_BUILD; then
         -Bcmake-out .
 
     echo "Building any Cadence-specific binaries on top"
-    CXXFLAGS="-fno-exceptions -fno-rtti" cmake -DBUCK2="$BUCK" \
+    CXXFLAGS="-fno-exceptions -fno-rtti" cmake \
         -DCMAKE_TOOLCHAIN_FILE=/home/zonglinpeng/ws/zonglinpeng/executorch/backends/cadence/cadence.cmake \
         -DCMAKE_INSTALL_PREFIX=cmake-out \
         -DCMAKE_BUILD_TYPE=Release \
@@ -57,7 +57,7 @@ if $STEPWISE_BUILD; then
 else
     echo "Building Cadence toolchain with ExecuTorch packages"
     cmake_prefix_path="${PWD}/cmake-out/lib/cmake/ExecuTorch;${PWD}/cmake-out/third-party/gflags"
-    CXXFLAGS="-fno-exceptions -fno-rtti" cmake -DBUCK2="$BUCK" \
+    CXXFLAGS="-fno-exceptions -fno-rtti" cmake \
         -DCMAKE_PREFIX_PATH="${cmake_prefix_path}" \
         -DHAVE_SYS_STAT_H=ON \
         -DCMAKE_TOOLCHAIN_FILE=./backends/cadence/cadence.cmake \

backends/cadence/build_cadence_hifi4.sh

@@ -35,7 +35,7 @@ if $STEPWISE_BUILD; then
         -Bcmake-out .
 
     echo "Building any Cadence-specific binaries on top"
-    CXXFLAGS="-fno-exceptions -fno-rtti" cmake -DBUCK2="$BUCK" \
+    CXXFLAGS="-fno-exceptions -fno-rtti" cmake \
         -DCMAKE_TOOLCHAIN_FILE=./backends/cadence/cadence.cmake \
         -DCMAKE_INSTALL_PREFIX=cmake-out \
         -DCMAKE_BUILD_TYPE=Release \
@@ -56,7 +56,7 @@ if $STEPWISE_BUILD; then
 else
     echo "Building Cadence toolchain with ExecuTorch packages"
     cmake_prefix_path="${PWD}/cmake-out/lib/cmake/ExecuTorch;${PWD}/cmake-out/third-party/gflags"
-    CXXFLAGS="-fno-exceptions -fno-rtti" cmake -DBUCK2="$BUCK" \
+    CXXFLAGS="-fno-exceptions -fno-rtti" cmake \
         -DCMAKE_PREFIX_PATH="${cmake_prefix_path}" \
         -DCMAKE_TOOLCHAIN_FILE=./backends/cadence/cadence.cmake \
         -DCMAKE_INSTALL_PREFIX=cmake-out \

backends/vulkan/runtime/graph/ComputeGraph.cpp

@@ -861,25 +861,6 @@ void ComputeGraph::prepare() {
   }
 
   execute_threshold_node_count_ = count_threshold;
-
-  for (SharedObject& shared_object : shared_objects_) {
-    shared_object.allocate(this);
-    shared_object.bind_users(this);
-  }
-}
-
-void ComputeGraph::prepare_pipelines() {
-  for (std::unique_ptr<PrepackNode>& node : prepack_nodes_) {
-    node->prepare_pipelines(this);
-  }
-  for (std::unique_ptr<ExecuteNode>& node : execute_nodes_) {
-    node->prepare_pipelines(this);
-  }
-  context_->pipeline_cache().create_pipelines(pipeline_descriptors_);
-
-  pipeline_descriptors_ = std::unordered_set<
-      vkapi::ComputePipelineCache::Key,
-      vkapi::ComputePipelineCache::Hasher>();
 }
 
 void ComputeGraph::prepare_pipelines() {

backends/xnnpack/partition/config/xnnpack_config.py

@@ -10,6 +10,11 @@
 from typing import List, Optional
 
 import torch
+from executorch.backends.xnnpack.utils.quant_utils import (
+    is_dequant,
+    is_qparam,
+    is_quant,
+)
 from executorch.exir.backend.canonical_partitioners.config_partitioner import (
     format_target_name,
     PartitionerConfig,
@@ -223,9 +228,18 @@ def _check_node_has_valid_dtype(self, node):
         valid_dtypes = {
             torch.float32,
             torch.float16,
-            torch.int8,
-            torch.qint8,
         }
+        # Only allow int8 and quant dtypes for quant operations
+        if is_quant(node) or is_dequant(node) or is_qparam(node):
+            valid_dtypes.update(
+                {
+                    torch.qint32,
+                    torch.qint8,
+                    torch.quint8,
+                    torch.int8,
+                }
+            )
+
         if (
             node.op != "placeholder"
             and node.op != "call_function"

examples/apple/mps/scripts/build_mps_executor_runner.sh

@@ -38,7 +38,7 @@ done
 
 rm -rf "$OUTPUT"
 
-cmake -DBUCK2="$BUCK" \
+cmake \
           -DCMAKE_INSTALL_PREFIX=cmake-out \
           -DCMAKE_BUILD_TYPE="$MODE" \
           -DEXECUTORCH_BUILD_DEVTOOLS=ON \

examples/apple/mps/test_mps.sh

@@ -15,7 +15,7 @@ cmake_install_executorch_devtools_lib() {
   echo "Installing libexecutorch.a, libportable_kernels.a, libetdump.a, libbundled_program.a"
   rm -rf cmake-out
 
-  retry cmake -DBUCK2="$BUCK" \
+  retry cmake \
           -DCMAKE_INSTALL_PREFIX=cmake-out \
           -DCMAKE_BUILD_TYPE=Release \
           -DEXECUTORCH_BUILD_DEVTOOLS=ON \
@@ -56,11 +56,5 @@ then
   PYTHON_EXECUTABLE=python3
 fi
 
-if [[ -z $BUCK ]];
-then
-  BUCK=buck2
-fi
-
-
 cmake_install_executorch_devtools_lib
 test_cmake_mps