Add graph breaks

.ci/docker/aotriton_version.txt

@@ -1,5 +1,5 @@
 0.6b
 manylinux_2_17
-rocm6.1
-7f07e8a1cb1f99627eb6d77f5c0e9295c775f3c7
-77c29fa3f3b614e187d7213d745e989a92708cee2bc6020419ab49019af399d1
+rocm6
+04b5df8c8123f90cba3ede7e971e6fbc6040d506
+3db6ecbc915893ff967abd6e1b43bd5f54949868873be60dc802086c3863e648

.ci/docker/ci_commit_pins/executorch.txt

@@ -1 +1 @@
-c572f9e509b5ec5d56f4d218271e36269bba244f
+172574a6be5910a4609e4ed1bef2b6b8475ddb3d

.ci/docker/ci_commit_pins/triton-rocm.txt

@@ -1 +1 @@
-21eae954efa5bf584da70324b640288c3ee7aede
+01cbe5045a6898c9a925f01435c8277b2fe6afcc

.ci/docker/ci_commit_pins/triton-xpu.txt

@@ -1 +1 @@
-1b2f15840e0d70eec50d84c7a0575cb835524def
+aac14a3b93f11d781d1d5ebc5400b15ae8df5185

.ci/docker/ci_commit_pins/triton.txt

@@ -1 +1 @@
-dedb7bdf339a3546896d4820366ca562c586bfa0
+45fff310c891f5a92d55445adf8cc9d29df5841e

.ci/docker/common/install_aotriton.sh

@@ -9,7 +9,7 @@ TARBALL='aotriton.tar.bz2'
 read -d "\n" VER MANYLINUX ROCMBASE PINNED_COMMIT SHA256 < aotriton_version.txt || true
 ARCH=$(uname -m)
 AOTRITON_INSTALL_PREFIX="$1"
-AOTRITON_URL="https://github.com/ROCm/aotriton/releases/download/${VER}/aotriton-${VER}-${MANYLINUX}_${ARCH}-${ROCMBASE}-shared.tar.bz2"
+AOTRITON_URL="https://github.com/ROCm/aotriton/releases/download/${VER}/aotriton-${VER}-${MANYLINUX}_${ARCH}-${ROCMBASE}.tar.bz2"
 
 cd "${AOTRITON_INSTALL_PREFIX}"
 # Must use -L to follow redirects

.ci/docker/common/install_conda.sh

@@ -85,7 +85,7 @@ fi
   else
     CONDA_COMMON_DEPS="astunparse pyyaml mkl=2021.4.0 mkl-include=2021.4.0 setuptools"
 
-    if [ "$ANACONDA_PYTHON_VERSION" = "3.11" ] || [ "$ANACONDA_PYTHON_VERSION" = "3.12" ] || [ "$ANACONDA_PYTHON_VERSION" = "3.13" ]; then
+    if [ "$ANACONDA_PYTHON_VERSION" = "3.11" ] || [ "$ANACONDA_PYTHON_VERSION" = "3.12" ]; then
       conda_install numpy=1.26.0 ${CONDA_COMMON_DEPS}
     else
       conda_install numpy=1.21.2 ${CONDA_COMMON_DEPS}

.ci/docker/requirements-ci.txt

@@ -306,7 +306,7 @@ pywavelets==1.5.0 ; python_version >= "3.12"
 #Pinned versions: 1.4.1
 #test that import:
 
-lxml==5.0.0
+lxml==5.0.0.
 #Description: This is a requirement of unittest-xml-reporting
 
 # Python-3.9 binaries

.ci/pytorch/build.sh

@@ -230,10 +230,6 @@ if [[ "${BUILD_ENVIRONMENT}" != *android* && "${BUILD_ENVIRONMENT}" != *cuda* ]]
   export BUILD_STATIC_RUNTIME_BENCHMARK=ON
 fi
 
-if [[ "$BUILD_ENVIRONMENT" == *-debug* ]]; then
-  export CMAKE_BUILD_TYPE=RelWithAssert
-fi
-
 # Do not change workspace permissions for ROCm CI jobs
 # as it can leave workspace with bad permissions for cancelled jobs
 if [[ "$BUILD_ENVIRONMENT" != *rocm* ]]; then

.ci/pytorch/common_utils.sh

@@ -222,8 +222,6 @@ function checkout_install_torchbench() {
     # to install and test other models
     python install.py --continue_on_fail
   fi
-  echo "Print all dependencies after TorchBench is installed"
-  python -mpip freeze
   popd
 }
 

.ci/pytorch/multigpu-test.sh

@@ -18,9 +18,8 @@ time python test/run_test.py --verbose -i distributed/test_c10d_gloo
 time python test/run_test.py --verbose -i distributed/test_c10d_nccl
 time python test/run_test.py --verbose -i distributed/test_c10d_spawn_gloo
 time python test/run_test.py --verbose -i distributed/test_c10d_spawn_nccl
-time python test/run_test.py --verbose -i distributed/test_compute_comm_reordering
+time python test/run_test.py --verbose -i distributed/test_cuda_p2p
 time python test/run_test.py --verbose -i distributed/test_store
-time python test/run_test.py --verbose -i distributed/test_symmetric_memory
 time python test/run_test.py --verbose -i distributed/test_pg_wrapper
 time python test/run_test.py --verbose -i distributed/rpc/cuda/test_tensorpipe_agent
 # FSDP tests

.ci/pytorch/test.sh

@@ -249,7 +249,9 @@ fi
 # This tests that the debug asserts are working correctly.
 if [[ "$BUILD_ENVIRONMENT" == *-debug* ]]; then
     echo "We are in debug mode: $BUILD_ENVIRONMENT. Expect the python assertion to fail"
-    (cd test && ! get_exit_code python -c "import torch; torch._C._crash_if_debug_asserts_fail(424242)")
+    # TODO: Enable the check after we setup the build to run debug asserts without having
+    #       to do a full (and slow) debug build
+    # (cd test && ! get_exit_code python -c "import torch; torch._C._crash_if_debug_asserts_fail(424242)")
 elif [[ "$BUILD_ENVIRONMENT" != *-bazel-* ]]; then
     # Noop when debug is disabled. Skip bazel jobs because torch isn't available there yet.
     echo "We are not in debug mode: $BUILD_ENVIRONMENT. Expect the assertion to pass"
@@ -262,6 +264,18 @@ elif [[ $TEST_CONFIG == 'nogpu_AVX512' ]]; then
   export ATEN_CPU_CAPABILITY=avx2
 fi
 
+# temp workarounds for https://github.com/pytorch/pytorch/issues/126692, remove when fixed
+if [[ "$BUILD_ENVIRONMENT" != *-bazel-* ]]; then
+  pushd test
+  CUDA_VERSION=$(python -c "import torch; print(torch.version.cuda)")
+  if [ "$CUDA_VERSION" == "12.4" ]; then
+    ISCUDA124="cu124"
+  else
+    ISCUDA124=""
+  fi
+  popd
+fi
+
 test_python_legacy_jit() {
   time python test/run_test.py --include test_jit_legacy test_jit_fuser_legacy --verbose
   assert_git_not_dirty
@@ -336,31 +350,17 @@ test_inductor_distributed() {
   assert_git_not_dirty
 }
 
-test_inductor_shard() {
-  if [[ -z "$NUM_TEST_SHARDS" ]]; then
-    echo "NUM_TEST_SHARDS must be defined to run a Python test shard"
-    exit 1
-  fi
-
+test_inductor() {
   python tools/dynamo/verify_dynamo.py
-  python test/run_test.py --inductor \
-    --include test_modules test_ops test_ops_gradients test_torch \
-    --shard "$1" "$NUM_TEST_SHARDS" \
-    --verbose
-
+  python test/run_test.py --inductor --include test_modules test_ops test_ops_gradients test_torch --verbose
   # Do not add --inductor for the following inductor unit tests, otherwise we will fail because of nested dynamo state
-  python test/run_test.py \
-    --include inductor/test_torchinductor inductor/test_torchinductor_opinfo inductor/test_aot_inductor \
-    --shard "$1" "$NUM_TEST_SHARDS" \
-    --verbose
-}
+  python test/run_test.py --include inductor/test_torchinductor inductor/test_torchinductor_opinfo inductor/test_aot_inductor --verbose
 
-test_inductor_aoti() {
   # docker build uses bdist_wheel which does not work with test_aot_inductor
   # TODO: need a faster way to build
   if [[ "$BUILD_ENVIRONMENT" != *rocm* ]]; then
-    BUILD_AOT_INDUCTOR_TEST=1 python setup.py develop
-    CPP_TESTS_DIR="${BUILD_BIN_DIR}" LD_LIBRARY_PATH="${TORCH_LIB_DIR}" python test/run_test.py --cpp --verbose -i cpp/test_aoti_abi_check cpp/test_aoti_inference
+      BUILD_AOT_INDUCTOR_TEST=1 python setup.py develop
+      CPP_TESTS_DIR="${BUILD_BIN_DIR}" LD_LIBRARY_PATH="${TORCH_LIB_DIR}" python test/run_test.py --cpp --verbose -i cpp/test_aoti_abi_check cpp/test_aoti_inference
   fi
 }
 
@@ -379,7 +379,7 @@ test_inductor_cpp_wrapper_abi_compatible() {
     --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_training.csv"
   python benchmarks/dynamo/check_accuracy.py \
     --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_training.csv" \
-    --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_timm_training.csv"
+    --expected "benchmarks/dynamo/ci_expected_accuracy/${ISCUDA124}/inductor_timm_training.csv"
 }
 
 # "Global" flags for inductor benchmarking controlled by TEST_CONFIG
@@ -404,7 +404,7 @@ if [[ "${TEST_CONFIG}" == *dynamic* ]]; then
   DYNAMO_BENCHMARK_FLAGS+=(--dynamic-shapes --dynamic-batch-only)
 fi
 
-if [[ "${TEST_CONFIG}" == *cpu_inductor* || "${TEST_CONFIG}" == *cpu_aot_inductor* ]]; then
+if [[ "${TEST_CONFIG}" == *cpu_inductor* ]]; then
   DYNAMO_BENCHMARK_FLAGS+=(--device cpu)
 else
   DYNAMO_BENCHMARK_FLAGS+=(--device cuda)
@@ -529,10 +529,9 @@ test_single_dynamo_benchmark() {
     test_perf_for_dashboard "$suite" \
       "${DYNAMO_BENCHMARK_FLAGS[@]}" "$@" "${partition_flags[@]}"
   else
-    if [[ "${TEST_CONFIG}" == *aot_inductor* && "${TEST_CONFIG}" != *cpu_aot_inductor* ]]; then
+    if [[ "${TEST_CONFIG}" == *aot_inductor* ]]; then
       # Test AOTInductor with the ABI-compatible mode on CI
       # This can be removed once the ABI-compatible mode becomes default.
-      # For CPU device, we perfer non ABI-compatible mode on CI when testing AOTInductor.
       export TORCHINDUCTOR_ABI_COMPATIBLE=1
     fi
     python "benchmarks/dynamo/$suite.py" \
@@ -542,10 +541,10 @@ test_single_dynamo_benchmark() {
       --output "$TEST_REPORTS_DIR/${name}_${suite}.csv"
     python benchmarks/dynamo/check_accuracy.py \
       --actual "$TEST_REPORTS_DIR/${name}_$suite.csv" \
-      --expected "benchmarks/dynamo/ci_expected_accuracy/${TEST_CONFIG}_${name}.csv"
+      --expected "benchmarks/dynamo/ci_expected_accuracy/${ISCUDA124}/${TEST_CONFIG}_${name}.csv"
     python benchmarks/dynamo/check_graph_breaks.py \
       --actual "$TEST_REPORTS_DIR/${name}_$suite.csv" \
-      --expected "benchmarks/dynamo/ci_expected_accuracy/${TEST_CONFIG}_${name}.csv"
+      --expected "benchmarks/dynamo/ci_expected_accuracy/${ISCUDA124}/${TEST_CONFIG}_${name}.csv"
   fi
 }
 
@@ -573,15 +572,11 @@ test_dynamo_benchmark() {
   elif [[ "${TEST_CONFIG}" == *perf* ]]; then
     test_single_dynamo_benchmark "dashboard" "$suite" "$shard_id" "$@"
   else
-    if [[ "${TEST_CONFIG}" == *cpu_inductor* || "${TEST_CONFIG}" == *cpu_aot_inductor* ]]; then
-      local dt="float32"
-      if [[ "${TEST_CONFIG}" == *amp* ]]; then
-        dt="amp"
-      fi
+    if [[ "${TEST_CONFIG}" == *cpu_inductor* ]]; then
       if [[ "${TEST_CONFIG}" == *freezing* ]]; then
-        test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --"$dt" --freezing "$@"
+        test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --float32 --freezing "$@"
       else
-        test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --"$dt" "$@"
+        test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --float32 "$@"
       fi
     elif [[ "${TEST_CONFIG}" == *aot_inductor* ]]; then
       test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --bfloat16 "$@"
@@ -605,7 +600,7 @@ test_inductor_torchbench_smoketest_perf() {
     --bfloat16 --inference --inductor --only moco --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv"
   python benchmarks/dynamo/check_accuracy.py \
     --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv" \
-    --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_torchbench_inference.csv"
+    --expected "benchmarks/dynamo/ci_expected_accuracy/${ISCUDA124}/inductor_torchbench_inference.csv"
 
   python benchmarks/dynamo/torchbench.py --device cuda --performance --backend inductor --float16 --training \
     --batch-size-file "$(realpath benchmarks/dynamo/torchbench_models_list.txt)" --only hf_Bert \
@@ -620,8 +615,13 @@ test_inductor_torchbench_smoketest_perf() {
   # https://github.com/pytorch/pytorch/actions/runs/7158691360/job/19491437314,
   # and thus we lower its threshold to reduce flakiness. If this continues to be a problem,
   # we switch to use some other model.
-  # lowering threshold from 4.9 to 4.7 for cu124. Will bump it up after cuda 12.4.0->12.4.1 update
-  python benchmarks/dynamo/check_perf_csv.py -f "$TEST_REPORTS_DIR/inductor_inference_smoketest.csv" -t 4.7
+  # Use 4.7 for cuda 12.4, change back to 4.9 after fixing https://github.com/pytorch/pytorch/issues/126692
+  if [ "$CUDA_VERSION" == "12.4" ]; then
+    THRESHOLD=4.7
+  else
+    THRESHOLD=4.9
+  fi
+  python benchmarks/dynamo/check_perf_csv.py -f "$TEST_REPORTS_DIR/inductor_inference_smoketest.csv" -t $THRESHOLD
 
   # Check memory compression ratio for a few models
   for test in hf_Albert timm_vision_transformer; do
@@ -640,7 +640,7 @@ test_inductor_torchbench_smoketest_perf() {
       --only $test --output "$TEST_REPORTS_DIR/inductor_warm_start_smoketest_$test.csv"
     python benchmarks/dynamo/check_accuracy.py \
       --actual "$TEST_REPORTS_DIR/inductor_warm_start_smoketest_$test.csv" \
-      --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_huggingface_training.csv"
+      --expected "benchmarks/dynamo/ci_expected_accuracy/${ISCUDA124}/inductor_huggingface_training.csv"
   done
 }
 
@@ -1266,14 +1266,13 @@ elif [[ "${TEST_CONFIG}" == *timm* ]]; then
   id=$((SHARD_NUMBER-1))
   test_dynamo_benchmark timm_models "$id"
 elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
-  if [[ "${TEST_CONFIG}" == *cpu_inductor* || "${TEST_CONFIG}" == *cpu_aot_inductor* ]]; then
+  if [[ "${TEST_CONFIG}" == *cpu_inductor* ]]; then
     install_torchaudio cpu
   else
     install_torchaudio cuda
   fi
   install_torchtext
   install_torchvision
-  TORCH_CUDA_ARCH_LIST="8.0;8.6" pip_install git+https://github.com/pytorch/ao.git
   id=$((SHARD_NUMBER-1))
   # https://github.com/opencv/opencv-python/issues/885
   pip_install opencv-python==4.8.0.74
@@ -1292,7 +1291,7 @@ elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
     checkout_install_torchbench
     # Do this after checkout_install_torchbench to ensure we clobber any
     # nightlies that torchbench may pull in
-    if [[ "${TEST_CONFIG}" != *cpu_inductor* && "${TEST_CONFIG}" != *cpu_aot_inductor* ]]; then
+    if [[ "${TEST_CONFIG}" != *cpu_inductor* ]]; then
       install_torchrec_and_fbgemm
     fi
     PYTHONPATH=$(pwd)/torchbench test_dynamo_benchmark torchbench "$id"
@@ -1300,19 +1299,17 @@ elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
 elif [[ "${TEST_CONFIG}" == *inductor_cpp_wrapper_abi_compatible* ]]; then
   install_torchvision
   test_inductor_cpp_wrapper_abi_compatible
-elif [[ "${TEST_CONFIG}" == *inductor* ]]; then
+elif [[ "${TEST_CONFIG}" == *inductor* && "${SHARD_NUMBER}" == 1 ]]; then
   install_torchvision
-  test_inductor_shard "${SHARD_NUMBER}"
-  if [[ "${SHARD_NUMBER}" == 1 ]]; then
-    test_inductor_aoti
-    test_inductor_distributed
-  fi
-elif [[ "${TEST_CONFIG}" == *dynamo* ]]; then
+  test_inductor
+  test_inductor_distributed
+elif [[ "${TEST_CONFIG}" == *dynamo* && "${SHARD_NUMBER}" == 1 && $NUM_TEST_SHARDS -gt 1 ]]; then
+  install_torchvision
+  test_dynamo_shard 1
+  test_aten
+elif [[ "${TEST_CONFIG}" == *dynamo* && $SHARD_NUMBER -gt 1 && $NUM_TEST_SHARDS -gt 1 ]]; then
   install_torchvision
   test_dynamo_shard "${SHARD_NUMBER}"
-  if [[ "${SHARD_NUMBER}" == 1 ]]; then
-    test_aten
-  fi
 elif [[ "${BUILD_ENVIRONMENT}" == *rocm* && -n "$TESTS_TO_INCLUDE" ]]; then
   install_torchvision
   test_python_shard "$SHARD_NUMBER"

.circleci/scripts/binary_linux_test.sh

@@ -97,16 +97,8 @@ if [[ "$PACKAGE_TYPE" == conda ]]; then
   )
 elif [[ "$PACKAGE_TYPE" != libtorch ]]; then
   if [[ "\$BUILD_ENVIRONMENT" != *s390x* ]]; then
-    if [[ "$USE_SPLIT_BUILD" == "true" ]]; then
-      pkg_no_python="$(ls -1 /final_pkgs/torch_no_python* | sort |tail -1)"
-      pkg_torch="$(ls -1 /final_pkgs/torch-* | sort |tail -1)"
-      # todo: after folder is populated use the pypi_pkg channel instead
-      pip install "\$pkg_no_python" "\$pkg_torch" --index-url "https://download.pytorch.org/whl/\${CHANNEL}/${DESIRED_CUDA}_pypi_pkg"
-      retry pip install -q numpy protobuf typing-extensions
-    else
-      pip install "\$pkg" --index-url "https://download.pytorch.org/whl/\${CHANNEL}/${DESIRED_CUDA}"
-      retry pip install -q numpy protobuf typing-extensions
-    fi
+    pip install "\$pkg" --index-url "https://download.pytorch.org/whl/\${CHANNEL}/${DESIRED_CUDA}"
+    retry pip install -q numpy protobuf typing-extensions
   else
     pip install "\$pkg"
     retry pip install -q numpy protobuf typing-extensions
@@ -118,12 +110,6 @@ if [[ "$PACKAGE_TYPE" == libtorch ]]; then
   cd /tmp/libtorch
 fi
 
-if [[ "$GPU_ARCH_TYPE" == xpu ]]; then
-  # Workaround for __mkl_tmp_MOD unbound variable issue, refer https://github.com/pytorch/pytorch/issues/130543
-  set +u
-  source /opt/intel/oneapi/pytorch-gpu-dev-0.5/oneapi-vars.sh
-fi
-
 # Test the package
 /builder/check_binary.sh
 

.circleci/scripts/binary_populate_env.sh

@@ -100,18 +100,30 @@ if [[ "$PACKAGE_TYPE" =~ .*wheel.* && -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_B
     fi
 fi
 
-# Set triton via PYTORCH_EXTRA_INSTALL_REQUIREMENTS for triton xpu package
-if [[ "$PACKAGE_TYPE" =~ .*wheel.* && -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*xpu.* && $(uname) == "Linux" ]]; then
-    TRITON_REQUIREMENT="pytorch-triton-xpu==${TRITON_VERSION}"
-    if [[ -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*dev.* ]]; then
-        TRITON_SHORTHASH=$(cut -c1-10 $PYTORCH_ROOT/.ci/docker/ci_commit_pins/triton-xpu.txt)
-        TRITON_REQUIREMENT="pytorch-triton-xpu==${TRITON_VERSION}+${TRITON_SHORTHASH}"
-    fi
-    if [[ -z "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:-}" ]]; then
-        export PYTORCH_EXTRA_INSTALL_REQUIREMENTS="${TRITON_REQUIREMENT}"
-    else
-        export PYTORCH_EXTRA_INSTALL_REQUIREMENTS="${PYTORCH_EXTRA_INSTALL_REQUIREMENTS} | ${TRITON_REQUIREMENT}"
+JAVA_HOME=
+BUILD_JNI=OFF
+if [[ "$PACKAGE_TYPE" == libtorch ]]; then
+  POSSIBLE_JAVA_HOMES=()
+  POSSIBLE_JAVA_HOMES+=(/usr/local)
+  POSSIBLE_JAVA_HOMES+=(/usr/lib/jvm/java-8-openjdk-amd64)
+  POSSIBLE_JAVA_HOMES+=(/Library/Java/JavaVirtualMachines/*.jdk/Contents/Home)
+  # Add the Windows-specific JNI path
+  POSSIBLE_JAVA_HOMES+=("$PWD/pytorch/.circleci/windows-jni/")
+  for JH in "${POSSIBLE_JAVA_HOMES[@]}" ; do
+    if [[ -e "$JH/include/jni.h" ]] ; then
+      # Skip if we're not on Windows but haven't found a JAVA_HOME
+      if [[ "$JH" == "$PWD/pytorch/.circleci/windows-jni/" && "$OSTYPE" != "msys" ]] ; then
+        break
+      fi
+      echo "Found jni.h under $JH"
+      JAVA_HOME="$JH"
+      BUILD_JNI=ON
+      break
     fi
+  done
+  if [ -z "$JAVA_HOME" ]; then
+    echo "Did not find jni.h"
+  fi
 fi
 
 cat >"$envfile" <<EOL
@@ -124,7 +136,6 @@ export DESIRED_PYTHON="${DESIRED_PYTHON:-}"
 export DESIRED_CUDA="$DESIRED_CUDA"
 export LIBTORCH_VARIANT="${LIBTORCH_VARIANT:-}"
 export BUILD_PYTHONLESS="${BUILD_PYTHONLESS:-}"
-export USE_SPLIT_BUILD="${USE_SPLIT_BUILD:-}"
 if [[ "${OSTYPE}" == "msys" ]]; then
   export LIBTORCH_CONFIG="${LIBTORCH_CONFIG:-}"
   if [[ "${LIBTORCH_CONFIG:-}" == 'debug' ]]; then
@@ -148,6 +159,8 @@ export TORCH_CONDA_BUILD_FOLDER='pytorch-nightly'
 export ANACONDA_USER='pytorch'
 
 export USE_FBGEMM=1
+export JAVA_HOME=$JAVA_HOME
+export BUILD_JNI=$BUILD_JNI
 export PIP_UPLOAD_FOLDER="$PIP_UPLOAD_FOLDER"
 export DOCKER_IMAGE="$DOCKER_IMAGE"
 

.circleci/scripts/binary_upload.sh

@@ -25,15 +25,6 @@ if [[ "${DRY_RUN}" = "disabled" ]]; then
   AWS_S3_CP="aws s3 cp"
 fi
 
-if [[ "${USE_SPLIT_BUILD:-false}" == "true" ]]; then
-  UPLOAD_SUBFOLDER="${UPLOAD_SUBFOLDER}_pypi_pkg"
-fi
-
-# this is special build with all dependencies packaged
-if [[ ${BUILD_NAME} == *-full* ]]; then
-  UPLOAD_SUBFOLDER="${UPLOAD_SUBFOLDER}_full"
-fi
-
 # Sleep 2 minutes between retries for conda upload
 retry () {
   "$@"  || (sleep 5m && "$@") || (sleep 5m && "$@") || (sleep 5m && "$@") || (sleep 5m && "$@")

.git-blame-ignore-revs

@@ -40,7 +40,3 @@ e6ec0efaf87703c5f889cfc20b29be455885d58d
 a53cda1ddc15336dc1ff0ce1eff2a49cdc5f882e
 # 2024-01-02 clangformat: fused adam #116583
 9dc68d1aa9e554d09344a10fff69f7b50b2d23a0
-# 2024-06-28 enable UFMT in `torch/storage.py`
-d80939e5e9337e8078f11489afefec59fd42f93b
-# 2024-06-28 enable UFMT in `torch.utils.data`
-7cf0b90e49689d45be91aa539fdf54cf2ea8a9a3

.github/actionlint.yaml

@@ -47,5 +47,3 @@ self-hosted-runner:
     - macos-latest-xlarge
     - macos-13-xlarge
     - macos-14-xlarge
-    # Organization-wide Intel hosted XPU runners
-    - linux.idc.xpu

.github/actions/test-pytorch-binary/action.yml

@@ -26,7 +26,6 @@ runs:
           -e PYTORCH_FINAL_PACKAGE_DIR \
           -e PYTORCH_ROOT \
           -e SKIP_ALL_TESTS \
-          -e USE_SPLIT_BUILD \
           --tty \
           --detach \
           -v "${GITHUB_WORKSPACE}/pytorch:/pytorch" \
@@ -36,8 +35,7 @@ runs:
           "${DOCKER_IMAGE}"
         )
 
-        echo "CONTAINER_NAME=${container_name}" >> "$GITHUB_ENV"
-        if [[ "${GPU_ARCH_TYPE}" != "rocm" && "${BUILD_ENVIRONMENT}" != "linux-aarch64-binary-manywheel" && "${BUILD_ENVIRONMENT}" != "linux-s390x-binary-manywheel" && "${GPU_ARCH_TYPE}" != "xpu" ]]; then
+        if [[ "${GPU_ARCH_TYPE}" != "rocm" && "${BUILD_ENVIRONMENT}" != "linux-aarch64-binary-manywheel" && "${BUILD_ENVIRONMENT}" != "linux-s390x-binary-manywheel" ]]; then
           # Propagate download.pytorch.org IP to container. This is only needed on Linux non aarch64 runner
           grep download.pytorch.org /etc/hosts | docker exec -i "${container_name}" bash -c "/bin/cat >> /etc/hosts"
         fi
@@ -48,9 +46,10 @@ runs:
         docker exec -t "${container_name}" bash -c "source ${BINARY_ENV_FILE} && bash -x /run.sh"
 
     - name: Cleanup docker
-      if: always() && (env.BUILD_ENVIRONMENT == 'linux-s390x-binary-manywheel' || env.GPU_ARCH_TYPE == 'xpu')
+      if: always() && env.BUILD_ENVIRONMENT == 'linux-s390x-binary-manywheel'
       shell: bash
       run: |
-        # on s390x or xpu stop the container for clean worker stop
+        # on s390x stop the container for clean worker stop
+        # ignore expansion of "docker ps -q" since it could be empty
         # shellcheck disable=SC2046
-        docker stop "${{ env.CONTAINER_NAME }}" || true
+        docker stop $(docker ps -q) || true

.github/ci_commit_pins/audio.txt

@@ -1 +1 @@
-69b2a0adc2ec03ab99990d7e8be3d4510438c148
+b829e936f7cc61b48149f5f957a451a38bf2a178

.github/merge_rules.yaml

@@ -286,7 +286,6 @@
   - test/cpp/dist_autograd/**
   - test/cpp/rpc/**
   approved_by:
-  - wconstab
   - mrshenli
   - pritamdamania87
   - zhaojuanmao
@@ -313,25 +312,6 @@
   - Lint
   - pull
 
-- name: DCP
-  patterns:
-  - torch/distributed/checkpoint/**
-  approved_by:
-  - LucasLLC
-  - fegin
-  - wz337
-  - saumishr
-  - daulet-askarov
-  - pradeepdfb
-  - kirtiteja
-  - mhorowitz
-  - saiteja64
-  mandatory_checks_name:
-  - EasyCLA
-  - Lint
-  - pull
-
-
 - name: IDEEP
   patterns:
   - third_party/ideep
@@ -395,21 +375,13 @@
 
 - name: CPU inductor
   patterns:
-  - torch/_inductor/mkldnn_ir.py
   - torch/_inductor/mkldnn_lowerings.py
   - torch/_inductor/fx_passes/mkldnn_fusion.py
   - torch/_inductor/fx_passes/quantization.py
-  - torch/_inductor/codegen/cpp_prefix.h
   - torch/_inductor/codegen/cpp.py
-  - torch/_inductor/codegen/cpp_utils.py
-  - torch/_inductor/codegen/cpp_micro_gemm.py
-  - torch/_inductor/codegen/cpp_template_kernel.py
-  - torch/_inductor/codegen/cpp_template.py
-  - torch/_inductor/codegen/cpp_gemm_template.py
   - test/inductor/test_mkldnn_pattern_matcher.py
-  - test/inductor/test_cpu_repro.py
+  - test/inductor/test_cpu_repo.py
   - test/inductor/test_cpu_cpp_wrapper.py
-  - test/inductor/test_cpu_select_algorithm.py
   - aten/src/ATen/cpu/**
   - aten/src/ATen/native/quantized/cpu/**
   - test/quantization/core/test_quantized_op.py

.github/scripts/amd/package_triton_wheel.sh

@@ -93,8 +93,6 @@ done
 
 # Copy Include Files
 cp -r $ROCM_HOME/include/hip $TRITON_ROCM_DIR/include
-cp -r $ROCM_HOME/include/roctracer $TRITON_ROCM_DIR/include
-cp -r $ROCM_HOME/include/hsa $TRITON_ROCM_DIR/include
 
 # Copy linker
 mkdir -p $TRITON_ROCM_DIR/llvm/bin

.github/scripts/build_triton_wheel.py

@@ -11,12 +11,8 @@ SCRIPT_DIR = Path(__file__).parent
 REPO_DIR = SCRIPT_DIR.parent.parent
 
 
-def read_triton_pin(device: str = "cuda") -> str:
-    triton_file = "triton.txt"
-    if device == "rocm":
-        triton_file = "triton-rocm.txt"
-    elif device == "xpu":
-        triton_file = "triton-xpu.txt"
+def read_triton_pin(rocm_hash: bool = False) -> str:
+    triton_file = "triton.txt" if not rocm_hash else "triton-rocm.txt"
     with open(REPO_DIR / ".ci" / "docker" / "ci_commit_pins" / triton_file) as f:
         return f.read().strip()
 
@@ -53,7 +49,7 @@ def build_triton(
     version: str,
     commit_hash: str,
     build_conda: bool = False,
-    device: str = "cuda",
+    build_rocm: bool = False,
     py_version: Optional[str] = None,
     release: bool = False,
 ) -> Path:
@@ -73,14 +69,11 @@ def build_triton(
         triton_basedir = Path(tmpdir) / "triton"
         triton_pythondir = triton_basedir / "python"
         triton_repo = "https://github.com/openai/triton"
-        if device == "rocm":
+        if build_rocm:
             triton_pkg_name = "pytorch-triton-rocm"
-        elif device == "xpu":
-            triton_pkg_name = "pytorch-triton-xpu"
-            triton_repo = "https://github.com/intel/intel-xpu-backend-for-triton"
         else:
             triton_pkg_name = "pytorch-triton"
-        check_call(["git", "clone", triton_repo, "triton"], cwd=tmpdir)
+        check_call(["git", "clone", triton_repo], cwd=tmpdir)
         if release:
             ver, rev, patch = version.split(".")
             check_call(
@@ -147,7 +140,7 @@ def build_triton(
             expected_version=None,
         )
 
-        if device == "rocm":
+        if build_rocm:
             check_call(
                 [f"{SCRIPT_DIR}/amd/package_triton_wheel.sh"],
                 cwd=triton_basedir,
@@ -162,7 +155,7 @@ def build_triton(
         whl_path = next(iter((triton_pythondir / "dist").glob("*.whl")))
         shutil.copy(whl_path, Path.cwd())
 
-        if device == "rocm":
+        if build_rocm:
             check_call(
                 [f"{SCRIPT_DIR}/amd/patch_triton_wheel.sh", Path.cwd()],
                 cwd=triton_basedir,
@@ -177,19 +170,17 @@ def main() -> None:
     parser = ArgumentParser("Build Triton binaries")
     parser.add_argument("--release", action="store_true")
     parser.add_argument("--build-conda", action="store_true")
-    parser.add_argument(
-        "--device", type=str, default="cuda", choices=["cuda", "rocm", "xpu"]
-    )
+    parser.add_argument("--build-rocm", action="store_true")
     parser.add_argument("--py-version", type=str)
     parser.add_argument("--commit-hash", type=str)
     parser.add_argument("--triton-version", type=str, default=read_triton_version())
     args = parser.parse_args()
 
     build_triton(
-        device=args.device,
+        build_rocm=args.build_rocm,
         commit_hash=args.commit_hash
         if args.commit_hash
-        else read_triton_pin(args.device),
+        else read_triton_pin(args.build_rocm),
         version=args.triton_version,
         build_conda=args.build_conda,
         py_version=args.py_version,

.github/scripts/generate_binary_build_matrix.py

@@ -8,7 +8,6 @@ architectures:
     * CPU
     * Latest CUDA
     * Latest ROCM
-    * Latest XPU
 """
 
 import os
@@ -25,7 +24,6 @@ CUDA_ARCHES_CUDNN_VERSION = {"11.8": "9", "12.1": "9", "12.4": "9"}
 
 ROCM_ARCHES = ["6.0", "6.1"]
 
-XPU_ARCHES = ["xpu"]
 
 CPU_CXX11_ABI_ARCH = ["cpu-cxx11-abi"]
 
@@ -50,7 +48,7 @@ PYTORCH_EXTRA_INSTALL_REQUIREMENTS = {
         "nvidia-curand-cu11==10.3.0.86; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusolver-cu11==11.4.1.48; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusparse-cu11==11.7.5.86; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nccl-cu11==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-nccl-cu11==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvtx-cu11==11.8.86; platform_system == 'Linux' and platform_machine == 'x86_64'"
     ),
     "12.1": (
@@ -63,7 +61,7 @@ PYTORCH_EXTRA_INSTALL_REQUIREMENTS = {
         "nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'"
     ),
     "12.4": (
@@ -76,7 +74,7 @@ PYTORCH_EXTRA_INSTALL_REQUIREMENTS = {
         "nvidia-curand-cu12==10.3.5.119; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusolver-cu12==11.6.0.99; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusparse-cu12==12.3.0.142; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvtx-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvjitlink-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64'"
     ),
@@ -134,8 +132,6 @@ def arch_type(arch_version: str) -> str:
         return "cuda"
     elif arch_version in ROCM_ARCHES:
         return "rocm"
-    elif arch_version in XPU_ARCHES:
-        return "xpu"
     elif arch_version in CPU_CXX11_ABI_ARCH:
         return "cpu-cxx11-abi"
     elif arch_version in CPU_AARCH64_ARCH:
@@ -160,7 +156,6 @@ WHEEL_CONTAINER_IMAGES = {
         gpu_arch: f"pytorch/manylinux-builder:rocm{gpu_arch}-{DEFAULT_TAG}"
         for gpu_arch in ROCM_ARCHES
     },
-    "xpu": f"pytorch/manylinux2_28-builder:xpu-{DEFAULT_TAG}",
     "cpu": f"pytorch/manylinux-builder:cpu-{DEFAULT_TAG}",
     "cpu-cxx11-abi": f"pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-{DEFAULT_TAG}",
     "cpu-aarch64": f"pytorch/manylinuxaarch64-builder:cpu-aarch64-{DEFAULT_TAG}",
@@ -226,7 +221,6 @@ def translate_desired_cuda(gpu_arch_type: str, gpu_arch_version: str) -> str:
         "cuda": f"cu{gpu_arch_version.replace('.', '')}",
         "cuda-aarch64": "cu124",
         "rocm": f"rocm{gpu_arch_version}",
-        "xpu": "xpu",
     }.get(gpu_arch_type, gpu_arch_version)
 
 
@@ -331,13 +325,13 @@ def generate_wheels_matrix(
         package_type = "manywheel"
 
     if python_versions is None:
-        python_versions = FULL_PYTHON_VERSIONS + ["3.13"]
+        python_versions = FULL_PYTHON_VERSIONS
 
     if arches is None:
         # Define default compute archivectures
         arches = ["cpu"]
         if os == "linux":
-            arches += CPU_CXX11_ABI_ARCH + CUDA_ARCHES + ROCM_ARCHES + XPU_ARCHES
+            arches += CPU_CXX11_ABI_ARCH + CUDA_ARCHES + ROCM_ARCHES
         elif os == "windows":
             arches += CUDA_ARCHES
         elif os == "linux-aarch64":
@@ -360,16 +354,9 @@ def generate_wheels_matrix(
                 or arch_version == "cpu-aarch64"
                 or arch_version == "cpu-s390x"
                 or arch_version == "cuda-aarch64"
-                or arch_version == "xpu"
                 else arch_version
             )
 
-            # TODO: Enable python 3.13 on rocm, xpu, aarch64, windows
-            if (
-                gpu_arch_type in ["rocm", "xpu"] or os != "linux"
-            ) and python_version == "3.13":
-                continue
-
             # 12.1 linux wheels require PYTORCH_EXTRA_INSTALL_REQUIREMENTS to install
             if (
                 arch_version in ["12.4", "12.1", "11.8"]
@@ -399,49 +386,6 @@ def generate_wheels_matrix(
                         ),
                     }
                 )
-                if arch_version != "cuda-aarch64":
-                    ret.append(
-                        {
-                            "python_version": python_version,
-                            "gpu_arch_type": gpu_arch_type,
-                            "gpu_arch_version": gpu_arch_version,
-                            "desired_cuda": translate_desired_cuda(
-                                gpu_arch_type, gpu_arch_version
-                            ),
-                            "use_split_build": "True",
-                            "devtoolset": "",
-                            "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
-                            "package_type": package_type,
-                            "pytorch_extra_install_requirements": (
-                                PYTORCH_EXTRA_INSTALL_REQUIREMENTS[arch_version]  # fmt: skip
-                                if os != "linux-aarch64"
-                                else ""
-                            ),
-                            "build_name": f"{package_type}-py{python_version}-{gpu_arch_type}{gpu_arch_version}-split".replace(  # noqa: B950
-                                ".", "_"
-                            ),
-                        }
-                    )
-                    # Special build building to use on Colab. PyThon 3.10 for 12.1 CUDA
-                    if python_version == "3.10" and arch_version == "12.1":
-                        ret.append(
-                            {
-                                "python_version": python_version,
-                                "gpu_arch_type": gpu_arch_type,
-                                "gpu_arch_version": gpu_arch_version,
-                                "desired_cuda": translate_desired_cuda(
-                                    gpu_arch_type, gpu_arch_version
-                                ),
-                                "use_split_build": "False",
-                                "devtoolset": "",
-                                "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
-                                "package_type": package_type,
-                                "pytorch_extra_install_requirements": "",
-                                "build_name": f"{package_type}-py{python_version}-{gpu_arch_type}{gpu_arch_version}-full".replace(  # noqa: B950
-                                    ".", "_"
-                                ),
-                            }
-                        )
             else:
                 ret.append(
                     {
@@ -452,9 +396,7 @@ def generate_wheels_matrix(
                             gpu_arch_type, gpu_arch_version
                         ),
                         "devtoolset": (
-                            "cxx11-abi"
-                            if arch_version in ["cpu-cxx11-abi", "xpu"]
-                            else ""
+                            "cxx11-abi" if arch_version == "cpu-cxx11-abi" else ""
                         ),
                         "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
                         "package_type": package_type,

.github/scripts/get_workflow_type.py

@@ -0,0 +1,103 @@
+import json
+from argparse import ArgumentParser
+from typing import Any, Tuple
+
+from github import Auth, Github
+from github.Issue import Issue
+
+
+WORKFLOW_LABEL_META = ""  # use meta runners
+WORKFLOW_LABEL_LF = "lf."  # use runners from the linux foundation
+LABEL_TYPE_KEY = "label_type"
+MESSAGE_KEY = "message"
+MESSAGE = ""  # Debug message to return to the caller
+
+
+def parse_args() -> Any:
+    parser = ArgumentParser("Get dynamic rollout settings")
+    parser.add_argument("--github-token", type=str, required=True, help="GitHub token")
+    parser.add_argument(
+        "--github-repo",
+        type=str,
+        required=False,
+        default="pytorch/test-infra",
+        help="GitHub repo to get the issue",
+    )
+    parser.add_argument(
+        "--github-issue", type=int, required=True, help="GitHub issue umber"
+    )
+    parser.add_argument(
+        "--github-user", type=str, required=True, help="GitHub username"
+    )
+    parser.add_argument(
+        "--github-branch", type=str, required=True, help="Current GitHub branch"
+    )
+
+    return parser.parse_args()
+
+
+def get_gh_client(github_token: str) -> Github:
+    auth = Auth.Token(github_token)
+    return Github(auth=auth)
+
+
+def get_issue(gh: Github, repo: str, issue_num: int) -> Issue:
+    repo = gh.get_repo(repo)
+    return repo.get_issue(number=issue_num)
+
+
+def is_exception_branch(branch: str) -> bool:
+    return branch.split("/")[0] in {"main", "nightly", "release", "landchecks"}
+
+
+def get_workflow_type(issue: Issue, username: str) -> Tuple[str, str]:
+    try:
+        user_list = issue.get_comments()[0].body.split()
+
+        if user_list[0] == "!":
+            MESSAGE = "LF Workflows are disabled for everyone. Using meta runners."
+            return WORKFLOW_LABEL_META, MESSAGE
+        elif user_list[0] == "*":
+            MESSAGE = "LF Workflows are enabled for everyone. Using LF runners."
+            return WORKFLOW_LABEL_LF, MESSAGE
+        elif username in user_list:
+            MESSAGE = f"LF Workflows are enabled for {username}. Using LF runners."
+            return WORKFLOW_LABEL_LF, MESSAGE
+        else:
+            MESSAGE = f"LF Workflows are disabled for {username}. Using meta runners."
+            return WORKFLOW_LABEL_META, MESSAGE
+    except Exception as e:
+        MESSAGE = f"Failed to get determine workflow type. Falling back to meta runners. Exception: {e}"
+        return WORKFLOW_LABEL_META, MESSAGE
+
+
+def main() -> None:
+    args = parse_args()
+
+    if is_exception_branch(args.github_branch):
+        output = {
+            LABEL_TYPE_KEY: WORKFLOW_LABEL_META,
+            MESSAGE_KEY: f"Exception branch: '{args.github_branch}', using meta runners",
+        }
+    else:
+        try:
+            gh = get_gh_client(args.github_token)
+            # The default issue we use - https://github.com/pytorch/test-infra/issues/5132
+            issue = get_issue(gh, args.github_repo, args.github_issue)
+            label_type, message = get_workflow_type(issue, args.github_user)
+            output = {
+                LABEL_TYPE_KEY: label_type,
+                MESSAGE_KEY: message,
+            }
+        except Exception as e:
+            output = {
+                LABEL_TYPE_KEY: WORKFLOW_LABEL_META,
+                MESSAGE_KEY: f"Failed to get issue. Falling back to meta runners. Exception: {e}",
+            }
+
+    json_output = json.dumps(output)
+    print(json_output)
+
+
+if __name__ == "__main__":
+    main()

.github/scripts/runner_determinator.py

@@ -1,210 +0,0 @@
-# flake8: noqa: G004
-
-import logging
-import os
-from argparse import ArgumentParser
-from logging import LogRecord
-from typing import Any, Iterable
-
-from github import Auth, Github
-from github.Issue import Issue
-
-
-WORKFLOW_LABEL_META = ""  # use meta runners
-WORKFLOW_LABEL_LF = "lf."  # use runners from the linux foundation
-
-GITHUB_OUTPUT = os.getenv("GITHUB_OUTPUT", "")
-GH_OUTPUT_KEY_LABEL_TYPE = "label-type"
-
-
-class ColorFormatter(logging.Formatter):
-    """Color codes the log messages based on the log level"""
-
-    COLORS = {
-        "WARNING": "\033[33m",  # Yellow
-        "ERROR": "\033[31m",  # Red
-        "CRITICAL": "\033[31m",  # Red
-        "INFO": "\033[0m",  # Reset
-        "DEBUG": "\033[0m",  # Reset
-    }
-
-    def format(self, record: LogRecord) -> str:
-        log_color = self.COLORS.get(record.levelname, "\033[0m")  # Default to reset
-        record.msg = f"{log_color}{record.msg}\033[0m"
-        return super().format(record)
-
-
-handler = logging.StreamHandler()
-handler.setFormatter(ColorFormatter(fmt="%(levelname)-8s: %(message)s"))
-
-log = logging.getLogger(os.path.basename(__file__))
-log.addHandler(handler)
-log.setLevel(logging.INFO)
-
-
-def set_github_output(key: str, value: str) -> None:
-    """
-    Defines outputs of the github action that invokes this script
-    """
-    if not GITHUB_OUTPUT:
-        # See https://github.blog/changelog/2022-10-11-github-actions-deprecating-save-state-and-set-output-commands/ for deprecation notice
-        log.warning(
-            "No env var found for GITHUB_OUTPUT, you must be running this code locally. Falling back to the deprecated print method."
-        )
-        print(f"::set-output name={key}::{value}")
-        return
-
-    with open(GITHUB_OUTPUT, "a") as f:
-        log.info(f"Setting output: {key}='{value}'")
-        f.write(f"{key}={value}\n")
-
-
-def parse_args() -> Any:
-    parser = ArgumentParser("Get dynamic rollout settings")
-    parser.add_argument("--github-token", type=str, required=True, help="GitHub token")
-    parser.add_argument(
-        "--github-issue-repo",
-        type=str,
-        required=False,
-        default="pytorch/test-infra",
-        help="GitHub repo to get the issue",
-    )
-    parser.add_argument(
-        "--github-repo",
-        type=str,
-        required=True,
-        help="GitHub repo where CI is running",
-    )
-    parser.add_argument(
-        "--github-issue", type=int, required=True, help="GitHub issue number"
-    )
-    parser.add_argument(
-        "--github-actor", type=str, required=True, help="GitHub triggering_actor"
-    )
-    parser.add_argument(
-        "--github-issue-owner", type=str, required=True, help="GitHub issue owner"
-    )
-    parser.add_argument(
-        "--github-branch", type=str, required=True, help="Current GitHub branch or tag"
-    )
-    parser.add_argument(
-        "--github-ref-type",
-        type=str,
-        required=True,
-        help="Current GitHub ref type, branch or tag",
-    )
-
-    return parser.parse_args()
-
-
-def get_gh_client(github_token: str) -> Github:
-    auth = Auth.Token(github_token)
-    return Github(auth=auth)
-
-
-def get_issue(gh: Github, repo: str, issue_num: int) -> Issue:
-    repo = gh.get_repo(repo)
-    return repo.get_issue(number=issue_num)
-
-
-def get_potential_pr_author(
-    gh: Github, repo: str, username: str, ref_type: str, ref_name: str
-) -> str:
-    # If the trigger was a new tag added by a bot, this is a ciflow case
-    # Fetch the actual username from the original PR. The PR number is
-    # embedded in the tag name: ciflow/<name>/<pr-number>
-    if username == "pytorch-bot[bot]" and ref_type == "tag":
-        split_tag = ref_name.split("/")
-        if (
-            len(split_tag) == 3
-            and split_tag[0] == "ciflow"
-            and split_tag[2].isnumeric()
-        ):
-            pr_number = split_tag[2]
-            try:
-                repository = gh.get_repo(repo)
-                pull = repository.get_pull(number=int(pr_number))
-            except Exception as e:
-                raise Exception(  # noqa: TRY002
-                    f"issue with pull request {pr_number} from repo {repository}"
-                ) from e
-            return pull.user.login
-    # In all other cases, return the original input username
-    return username
-
-
-def is_exception_branch(branch: str) -> bool:
-    return branch.split("/")[0] in {"main", "nightly", "release", "landchecks"}
-
-
-def get_workflow_type(issue: Issue, workflow_requestors: Iterable[str]) -> str:
-    try:
-        first_comment = issue.get_comments()[0].body.strip("\n\t ")
-
-        if first_comment[0] == "!":
-            log.info("LF Workflows are disabled for everyone. Using meta runners.")
-            return WORKFLOW_LABEL_META
-        elif first_comment[0] == "*":
-            log.info("LF Workflows are enabled for everyone. Using LF runners.")
-            return WORKFLOW_LABEL_LF
-        else:
-            all_opted_in_users = {
-                usr_raw.strip("\n\t@ ") for usr_raw in first_comment.split()
-            }
-            opted_in_requestors = {
-                usr for usr in workflow_requestors if usr in all_opted_in_users
-            }
-            if opted_in_requestors:
-                log.info(
-                    f"LF Workflows are enabled for {', '.join(opted_in_requestors)}. Using LF runners."
-                )
-                return WORKFLOW_LABEL_LF
-            else:
-                log.info(
-                    f"LF Workflows are disabled for {', '.join(workflow_requestors)}. Using meta runners."
-                )
-                return WORKFLOW_LABEL_META
-
-    except Exception as e:
-        log.error(
-            f"Failed to get determine workflow type. Falling back to meta runners. Exception: {e}"
-        )
-        return WORKFLOW_LABEL_META
-
-
-def main() -> None:
-    args = parse_args()
-
-    if args.github_ref_type == "branch" and is_exception_branch(args.github_branch):
-        log.info(f"Exception branch: '{args.github_branch}', using meta runners")
-        label_type = WORKFLOW_LABEL_META
-    else:
-        try:
-            gh = get_gh_client(args.github_token)
-            # The default issue we use - https://github.com/pytorch/test-infra/issues/5132
-            issue = get_issue(gh, args.github_issue_repo, args.github_issue)
-            username = get_potential_pr_author(
-                gh,
-                args.github_repo,
-                args.github_actor,
-                args.github_ref_type,
-                args.github_branch,
-            )
-            label_type = get_workflow_type(
-                issue,
-                (
-                    args.github_issue_owner,
-                    username,
-                ),
-            )
-        except Exception as e:
-            log.error(
-                f"Failed to get issue. Falling back to meta runners. Exception: {e}"
-            )
-            label_type = WORKFLOW_LABEL_META
-
-    set_github_output(GH_OUTPUT_KEY_LABEL_TYPE, label_type)
-
-
-if __name__ == "__main__":
-    main()

.github/scripts/sync_distributed_folder_prototype.sh

@@ -2,7 +2,7 @@
 
 set -eoux pipefail
 
-SYNC_BRANCH=pytorch-stable-prototype
+SYNC_BRANCH=fbcode/pytorch-stable-prototype
 
 git config user.email "fake@example.com"
 git config user.name  "PyTorch Stable Bot"
@@ -11,9 +11,7 @@ git fetch origin main
 git fetch origin "$SYNC_BRANCH"
 git checkout "$SYNC_BRANCH"
 
-# Using a hardcoded SHA here is a massive speedup as we can skip the entire history of the pytorch GitHub repo.
-# This specific SHA was chosen as it was before the "branch point" of the stable branch
-for SHA in $(git log ba3b05fdf37ddbc3c301294d6a560a816335e717..origin/main --pretty="%h" --reverse -- torch/distributed torch/csrc/distributed test/distributed test/cpp/c10d benchmarks/distributed)
+for SHA in $(git log 4333e122d4b74cdf84351ed2907045c6a767b4cd..origin/main --pretty="%h" --reverse -- torch/distributed torch/csrc/distributed test/distributed test/cpp/c10d benchmarks/distributed)
 do
     # `git merge-base --is-ancestor` exits with code 0 if the given SHA is an ancestor, and non-0 otherwise
     if git merge-base --is-ancestor $SHA HEAD || [[ $(git log --grep="(cherry picked from commit $SHA") ]]
@@ -22,12 +20,7 @@ do
         continue
     fi
     echo "Copying $SHA"
-    git cherry-pick -x "$SHA" -X theirs
-    git reset --soft HEAD~1
-    git add torch/distributed torch/csrc/distributed test/distributed test/cpp/c10d benchmarks/distributed
-    git checkout .
-    git commit --reuse-message=HEAD@{1}
-    git clean -f
+    git cherry-pick -x "$SHA"
 done
 
 if [[ "${WITH_PUSH}" == true ]]; then

.github/scripts/tag_docker_images_for_release.py

@@ -41,7 +41,7 @@ def main() -> None:
     )
 
     options = parser.parse_args()
-    tagged_images: Dict[str, bool] = {}
+    tagged_images: Dict[str, bool] = dict()
     platform_images = [
         generate_binary_build_matrix.WHEEL_CONTAINER_IMAGES,
         generate_binary_build_matrix.LIBTORCH_CONTAINER_IMAGES,

.github/scripts/td_llm_indexer.sh

@@ -7,7 +7,6 @@ cd llm-target-determinator
 pip install -q -r requirements.txt
 cd ../codellama
 pip install -e .
-pip install numpy==1.26.0
 
 # Run indexer
 cd ../llm-target-determinator

.github/scripts/test_trymerge.py

@@ -397,7 +397,6 @@ class TestTryMerge(TestCase):
         # self.assertGreater(len(pr.get_checkrun_conclusions()), 3)
         self.assertGreater(pr.get_commit_count(), 60)
 
-    @skip("GitHub doesn't keep this data anymore")
     def test_gql_retrieve_checksuites(self, *args: Any) -> None:
         "Fetch comments and conclusions for PR with 60 commits"
         pr = GitHubPR("pytorch", "pytorch", 94787)
@@ -895,24 +894,6 @@ class TestBypassFailures(TestCase):
         self.assertTrue(len(ignorable["FLAKY"]) == 1)
         self.assertTrue(len(ignorable["BROKEN_TRUNK"]) == 0)
 
-    def test_ignore_failures_older_run_same_workflow(self, *args: Any) -> None:
-        pr = GitHubPR("pytorch", "pytorch", 129013)
-        checks = pr.get_checkrun_conclusions()
-        checks = get_classifications(
-            pr.pr_num,
-            pr.project,
-            checks,
-            [],
-        )
-        pending, failed, ignorable = categorize_checks(
-            checks,
-            list(checks.keys()),
-        )
-        self.assertTrue(len(pending) == 0)
-        self.assertTrue(len(failed) == 0)
-        self.assertTrue(len(ignorable["FLAKY"]) == 2)
-        self.assertTrue(len(ignorable["UNSTABLE"]) == 13)
-
     @mock.patch("trymerge.read_merge_rules", side_effect=xla_merge_rules)
     def test_dont_ignore_flaky_failures(self, *args: Any) -> None:
         """
@@ -1041,7 +1022,7 @@ class TestGitHubPRGhstackDependencies(TestCase):
         )
 
     @skip(
-        reason="This test is run against a mutable PR that has changed, so it no longer works. The test should be changed"
+        reason="This test is run against a mutalbe PR that has changed, so it no longer works. The test should be changed"
     )
     @mock.patch("trymerge.read_merge_rules")
     @mock.patch("trymerge.GitRepo")

.github/scripts/trymerge.py

@@ -81,10 +81,9 @@ JobNameToStateDict = Dict[str, JobCheckState]
 
 
 class WorkflowCheckState:
-    def __init__(self, name: str, url: str, run_id: int, status: Optional[str]):
+    def __init__(self, name: str, url: str, status: Optional[str]):
         self.name: str = name
         self.url: str = url
-        self.run_id: int = run_id
         self.status: Optional[str] = status
         self.jobs: JobNameToStateDict = {}
 
@@ -123,7 +122,6 @@ fragment PRCheckSuites on CheckSuiteConnection {
       workflowRun {
         workflow {
           name
-          databaseId
         }
         databaseId
         url
@@ -514,7 +512,7 @@ def add_workflow_conclusions(
     workflows: Dict[str, WorkflowCheckState] = {}
 
     # for the jobs that don't have a workflow
-    no_workflow_obj: WorkflowCheckState = WorkflowCheckState("", "", 0, None)
+    no_workflow_obj: WorkflowCheckState = WorkflowCheckState("", "", None)
 
     def add_conclusions(edges: Any) -> None:
         for edge_idx, edge in enumerate(edges):
@@ -525,30 +523,18 @@ def add_workflow_conclusions(
             workflow_obj: WorkflowCheckState = no_workflow_obj
 
             if workflow_run is not None:
-                # This is the usual workflow run ID we see on GitHub
-                workflow_run_id = workflow_run["databaseId"]
-                # While this is the metadata name and ID of the workflow itself
                 workflow_name = workflow_run["workflow"]["name"]
-                workflow_id = workflow_run["workflow"]["databaseId"]
-
                 workflow_conclusion = node["conclusion"]
                 # Do not override existing status with cancelled
                 if workflow_conclusion == "CANCELLED" and workflow_name in workflows:
                     continue
-
-                # Only keep the latest workflow run for each workflow, heuristically,
-                # it's the run with largest run ID
-                if (
-                    workflow_id not in workflows
-                    or workflows[workflow_id].run_id < workflow_run_id
-                ):
-                    workflows[workflow_id] = WorkflowCheckState(
+                if workflow_name not in workflows:
+                    workflows[workflow_name] = WorkflowCheckState(
                         name=workflow_name,
                         status=workflow_conclusion,
                         url=workflow_run["url"],
-                        run_id=workflow_run_id,
                     )
-                workflow_obj = workflows[workflow_id]
+                workflow_obj = workflows[workflow_name]
 
             while checkruns is not None:
                 for checkrun_node in checkruns["nodes"]:
@@ -586,12 +572,12 @@ def add_workflow_conclusions(
     # the jobs in but don't put the workflow in.  We care more about the jobs in
     # the workflow that ran than the container workflow.
     res: JobNameToStateDict = {}
-    for workflow in workflows.values():
+    for workflow_name, workflow in workflows.items():
         if len(workflow.jobs) > 0:
             for job_name, job in workflow.jobs.items():
                 res[job_name] = job
         else:
-            res[workflow.name] = JobCheckState(
+            res[workflow_name] = JobCheckState(
                 workflow.name,
                 workflow.url,
                 workflow.status,
@@ -1177,6 +1163,7 @@ class GitHubPR:
             # Finally, upload the record to Rockset. The list of pending and failed
             # checks are at the time of the merge
             save_merge_record(
+                collection=ROCKSET_MERGES_COLLECTION,
                 comment_id=comment_id,
                 pr_num=self.pr_num,
                 owner=self.org,
@@ -1192,8 +1179,10 @@ class GitHubPR:
                 merge_base_sha=self.get_merge_base(),
                 merge_commit_sha=merge_commit_sha,
                 is_failed=False,
+                dry_run=dry_run,
                 skip_mandatory_checks=skip_mandatory_checks,
                 ignore_current=bool(ignore_current_checks),
+                workspace=ROCKSET_MERGES_WORKSPACE,
             )
         else:
             print("Missing comment ID or PR number, couldn't upload to Rockset")
@@ -1500,6 +1489,7 @@ def checks_to_markdown_bullets(
 
 @retries_decorator()
 def save_merge_record(
+    collection: str,
     comment_id: int,
     pr_num: int,
     owner: str,
@@ -1515,44 +1505,59 @@ def save_merge_record(
     merge_base_sha: str,
     merge_commit_sha: str = "",
     is_failed: bool = False,
+    dry_run: bool = False,
     skip_mandatory_checks: bool = False,
     ignore_current: bool = False,
     error: str = "",
+    workspace: str = "commons",
 ) -> None:
     """
-    This saves the merge records as a json, which can later be uploaded to s3
+    This saves the merge records into Rockset, so we can query them (for fun and profit)
     """
+    if dry_run:
+        # Decide not to save the record to Rockset if dry-run is set to not pollute
+        # the collection
+        return
 
-    # Prepare the record to be written into Rockset
-    data = [
-        {
-            "comment_id": comment_id,
-            "pr_num": pr_num,
-            "owner": owner,
-            "project": project,
-            "author": author,
-            "pending_checks": pending_checks,
-            "failed_checks": failed_checks,
-            "ignore_current_checks": ignore_current_checks,
-            "broken_trunk_checks": broken_trunk_checks,
-            "flaky_checks": flaky_checks,
-            "unstable_checks": unstable_checks,
-            "last_commit_sha": last_commit_sha,
-            "merge_base_sha": merge_base_sha,
-            "merge_commit_sha": merge_commit_sha,
-            "is_failed": is_failed,
-            "skip_mandatory_checks": skip_mandatory_checks,
-            "ignore_current": ignore_current,
-            "error": error,
-            # This is a unique identifier for the record for deduping purposes
-            # in rockset.  Any unique string would work
-            "_id": f"{project}-{pr_num}-{comment_id}-{os.environ.get('GITHUB_RUN_ID')}",
-        }
-    ]
-    repo_root = Path(__file__).resolve().parent.parent.parent
+    try:
+        import rockset  # type: ignore[import]
 
-    with open(repo_root / "merge_record.json", "w") as f:
-        json.dump(data, f)
+        # Prepare the record to be written into Rockset
+        data = [
+            {
+                "comment_id": comment_id,
+                "pr_num": pr_num,
+                "owner": owner,
+                "project": project,
+                "author": author,
+                "pending_checks": pending_checks,
+                "failed_checks": failed_checks,
+                "ignore_current_checks": ignore_current_checks,
+                "broken_trunk_checks": broken_trunk_checks,
+                "flaky_checks": flaky_checks,
+                "unstable_checks": unstable_checks,
+                "last_commit_sha": last_commit_sha,
+                "merge_base_sha": merge_base_sha,
+                "merge_commit_sha": merge_commit_sha,
+                "is_failed": is_failed,
+                "skip_mandatory_checks": skip_mandatory_checks,
+                "ignore_current": ignore_current,
+                "error": error,
+            }
+        ]
+
+        client = rockset.RocksetClient(
+            host="api.usw2a1.rockset.com", api_key=os.environ["ROCKSET_API_KEY"]
+        )
+        client.Documents.add_documents(
+            collection=collection,
+            data=data,
+            workspace=workspace,
+        )
+
+    except ModuleNotFoundError:
+        print("Rockset is missing, no record will be saved")
+        return
 
 
 @retries_decorator(rc=[])
@@ -2369,6 +2374,7 @@ def main() -> None:
             # list of pending and failed checks here, but they are not really
             # needed at the moment
             save_merge_record(
+                collection=ROCKSET_MERGES_COLLECTION,
                 comment_id=args.comment_id,
                 pr_num=args.pr_num,
                 owner=org,
@@ -2383,9 +2389,11 @@ def main() -> None:
                 last_commit_sha=pr.last_commit().get("oid", ""),
                 merge_base_sha=pr.get_merge_base(),
                 is_failed=True,
+                dry_run=args.dry_run,
                 skip_mandatory_checks=args.force,
                 ignore_current=args.ignore_current,
                 error=str(e),
+                workspace=ROCKSET_MERGES_WORKSPACE,
             )
         else:
             print("Missing comment ID or PR number, couldn't upload to Rockset")

.github/templates/linux_binary_build_workflow.yml.j2

@@ -81,7 +81,7 @@ jobs:
   !{{ config["build_name"] }}-test:  # Testing
     if: ${{ github.repository_owner == 'pytorch' }}
     needs: !{{ config["build_name"] }}-build
-    {%- if config["gpu_arch_type"] not in ["rocm", "xpu"] %}
+    {%- if config["gpu_arch_type"] != "rocm" %}
     uses: ./.github/workflows/_binary-test-linux.yml
     with:!{{ upload.binary_env_as_input(config) }}
       build_name: !{{ config["build_name"] }}
@@ -101,40 +101,6 @@ jobs:
       {%- endif %}
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
-    {%- elif config["gpu_arch_type"] == "xpu" %}
-    runs-on: linux.idc.xpu
-    timeout-minutes: !{{ common.timeout_minutes }}
-    !{{ upload.binary_env(config) }}
-    permissions:
-      id-token: write
-      contents: read
-    steps:
-      - name: Setup XPU
-        uses: ./.github/actions/setup-xpu
-      - name: configure aws credentials
-        id: aws_creds
-        uses: aws-actions/configure-aws-credentials@v1.7.0
-        with:
-          role-to-assume: arn:aws:iam::308535385114:role/gha_workflow_s3_and_ecr_read_only
-          aws-region: us-east-1
-      - name: Login to Amazon ECR
-        id: login-ecr
-        uses: aws-actions/amazon-ecr-login@v2
-      - uses: !{{ common.download_artifact_action }}
-        name: Download Build Artifacts
-        with:
-          name: !{{ config["build_name"] }}
-          path: "${{ runner.temp }}/artifacts/"
-      !{{ common.checkout(deep_clone=False, directory="pytorch") }}
-      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch) }}
-      - name: Pull Docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
-        with:
-          docker-image: !{{ config["container_image"] }}
-      - name: Test Pytorch binary
-        uses: ./pytorch/.github/actions/test-pytorch-binary
-      - name: Teardown XPU
-        uses: ./.github/actions/teardown-xpu
     {%- else %}
     runs-on: linux.rocm.gpu
     timeout-minutes: !{{ common.timeout_minutes }}

.github/templates/upload.yml.j2

@@ -30,9 +30,6 @@
   {%- if config["devtoolset"] %}
       DESIRED_DEVTOOLSET: !{{ config["devtoolset"] }}
   {%- endif %}
-  {%- if config.use_split_build is defined %}
-      use_split_build: !{{ config["use_split_build"] }}
-  {%- endif %}
 {%- endif %}
 {%- if config["package_type"] == "libtorch" %}
   {%- if config["libtorch_config"] %}
@@ -47,7 +44,6 @@
       # without this value pip does not get installed for some reason
       DESIRED_PYTHON: "3.8"
   {%- endif %}
-
 {%- else %}
       DESIRED_PYTHON: "!{{ config["python_version"] }}"
 {%- endif %}

.github/workflows/_bazel-build-test.yml

@@ -27,11 +27,6 @@ on:
         type: string
         description: |
           A JSON description of what configs to run later on.
-      runner:
-        required: false
-        type: string
-        default: "linux.large"
-        description: Runner type
 
 env:
   GIT_DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
@@ -39,7 +34,7 @@ env:
 jobs:
   filter:
     if: github.repository_owner == 'pytorch'
-    runs-on: ${{ inputs.runner }}
+    runs-on: [self-hosted, linux.large]
     outputs:
       test-matrix: ${{ steps.filter.outputs.test-matrix }}
       is-test-matrix-empty: ${{ steps.filter.outputs.is-test-matrix-empty }}

.github/workflows/_binary-build-linux.yml

@@ -21,13 +21,6 @@ on:
         default: 210
         type: number
         description: timeout for the job
-      use_split_build:
-        description: |
-          [Experimental] Build a libtorch only wheel and build pytorch such that
-          are built from the libtorch wheel.
-        required: false
-        type: boolean
-        default: false
       ALPINE_IMAGE:
         required: false
         type: string
@@ -117,7 +110,6 @@ jobs:
       PR_NUMBER: ${{ github.event.pull_request.number }}
       PYTORCH_FINAL_PACKAGE_DIR: /artifacts
       SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
-      USE_SPLIT_BUILD: ${{ inputs.use_split_build }}
     steps:
       - name: Make the env permanent during this workflow (but not the secrets)
         shell: bash
@@ -145,7 +137,6 @@ jobs:
             echo "PR_NUMBER=${{ env.PR_NUMBER }}"
             echo "PYTORCH_FINAL_PACKAGE_DIR=${{ env.PYTORCH_FINAL_PACKAGE_DIR }}"
             echo "SHA1=${{ env.SHA1 }}"
-            echo "USE_SPLIT_BUILD=${{ env.use_split_build }}"
           } >> "${GITHUB_ENV} }}"
 
       - name: List the env
@@ -255,7 +246,6 @@ jobs:
             -e PYTORCH_ROOT \
             -e SKIP_ALL_TESTS \
             -e PYTORCH_EXTRA_INSTALL_REQUIREMENTS \
-            -e USE_SPLIT_BUILD \
             --tty \
             --detach \
             -v "${GITHUB_WORKSPACE}/pytorch:/pytorch" \

.github/workflows/_binary-test-linux.yml

@@ -63,13 +63,6 @@ on:
         required: true
         type: string
         description: Hardware to run this job on. Valid values are linux.4xlarge, linux.4xlarge.nvidia.gpu, linux.arm64.2xlarge, and linux.rocm.gpu
-      use_split_build:
-        description: |
-          [Experimental] Build a libtorch only wheel and build pytorch such that
-          are built from the libtorch wheel.
-        required: false
-        type: boolean
-        default: false
     secrets:
       github-token:
         required: true
@@ -104,7 +97,6 @@ jobs:
       PR_NUMBER: ${{ github.event.pull_request.number }}
       PYTORCH_FINAL_PACKAGE_DIR: /artifacts
       SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
-      USE_SPLIT_BUILD: ${{ inputs.use_split_build }}
     steps:
       - name: Make the env permanent during this workflow (but not the secrets)
         shell: bash
@@ -132,7 +124,6 @@ jobs:
             echo "PR_NUMBER=${{ env.PR_NUMBER }}"
             echo "PYTORCH_FINAL_PACKAGE_DIR=${{ env.PYTORCH_FINAL_PACKAGE_DIR }}"
             echo "SHA1=${{ env.SHA1 }}"
-            echo "USE_SPLIT_BUILD=${{ env.USE_SPLIT_BUILD }}"
           } >> "${GITHUB_ENV} }}"
 
       - name: "[FB EMPLOYEES] Enable SSH (Click me for login details)"

.github/workflows/_binary-upload.yml

@@ -55,13 +55,6 @@ on:
         required: false
         type: string
         description: Desired python version
-      use_split_build:
-        description: |
-          [Experimental] Build a libtorch only wheel and build pytorch such that
-          are built from the libtorch wheel.
-        required: false
-        type: boolean
-        default: false
     secrets:
       github-token:
         required: true
@@ -100,7 +93,6 @@ jobs:
       PR_NUMBER: ${{ github.event.pull_request.number }}
       PYTORCH_FINAL_PACKAGE_DIR: /artifacts
       SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
-      USE_SPLIT_BUILD: ${{ inputs.use_split_build }}
     steps:
       - name: Checkout PyTorch
         uses: pytorch/pytorch/.github/actions/checkout-pytorch@main

.github/workflows/_linux-build-rg.yml

@@ -0,0 +1,105 @@
+name: linux-build-rg
+
+on:
+  workflow_call:
+    inputs:
+      build-environment:
+        required: true
+        type: string
+        description: Top-level label for what's being built/tested.
+      docker-image-name:
+        required: true
+        type: string
+        description: Name of the base docker image to build with.
+      build-generates-artifacts:
+        required: false
+        type: boolean
+        default: true
+        description: If set, upload generated build artifacts.
+      build-with-debug:
+        required: false
+        type: boolean
+        default: false
+        description: If set, build in debug mode.
+      sync-tag:
+        required: false
+        type: string
+        default: ""
+        description: |
+          If this is set, our linter will use this to make sure that every other
+          job with the same `sync-tag` is identical.
+      cuda-arch-list:
+        required: false
+        type: string
+        default: "5.2"
+        description: |
+          List of CUDA architectures CI build should target.
+      runner-group:
+        required: false
+        type: string
+        default: "arc-lf-linux.2xlarge"
+        description: Runner group to select group type
+      test-matrix:
+        required: false
+        type: string
+        description: |
+          An option JSON description of what test configs to run later on. This
+          is moved here from the Linux test workflow so that we can apply filter
+          logic using test-config labels earlier and skip unnecessary builds
+      s3-bucket:
+        description: S3 bucket to download artifact
+        required: false
+        type: string
+        default: "gha-artifacts"
+      aws-role-to-assume:
+        description: role to assume for downloading artifacts
+        required: false
+        type: string
+        default: ""
+    secrets:
+      HUGGING_FACE_HUB_TOKEN:
+        required: false
+        description: |
+          HF Auth token to avoid rate limits when downloading models or datasets from hub
+
+    outputs:
+      docker-image:
+        value: ${{ jobs.build.outputs.docker-image }}
+        description: The docker image containing the built PyTorch.
+      test-matrix:
+        value: ${{ jobs.build.outputs.test-matrix }}
+        description: An optional JSON description of what test configs to run later on.
+
+jobs:
+  build:
+    # Don't run on forked repos
+    if: github.repository_owner == 'pytorch'
+    runs-on:
+      group: ${{ inputs.runner-group }}
+    timeout-minutes: 240
+    outputs:
+      docker-image: ${{ steps.linux-build.outputs.docker-image }}
+      test-matrix: ${{ steps.linux-build.outputs.test-matrix }}
+    steps:
+      # [pytorch repo ref]
+      # Use a pytorch/pytorch reference instead of a reference to the local
+      # checkout because when we run this action we don't *have* a local
+      # checkout. In other cases you should prefer a local checkout.
+      - name: Checkout PyTorch
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+
+      - name: Linux Build
+        id: linux-build
+        uses: ./.github/actions/linux-build
+        with:
+          build-environment: ${{ inputs.build-environment }}
+          docker-image-name: ${{ inputs.docker-image-name }}
+          build-generates-artifacts: ${{ inputs.build-generates-artifacts }}
+          build-with-debug: ${{ inputs.build-with-debug }}
+          sync-tag: ${{ inputs.sync-tag }}
+          cuda-arch-list: ${{ inputs.cuda-arch-list }}
+          test-matrix: ${{ inputs.test-matrix }}
+          s3-bucket: ${{ inputs.s3-bucket }}
+          aws-role-to-assume: ${{ inputs.aws-role-to-assume }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}

.github/workflows/_linux-build.yml

@@ -39,7 +39,7 @@ on:
         type: string
         default: "linux.2xlarge"
         description: |
-          Label of the runner this job should run on.
+          List of CUDA architectures CI build should target.
       test-matrix:
         required: false
         type: string
@@ -228,7 +228,7 @@ jobs:
 
       - name: Store PyTorch Build Artifacts on S3
         uses: seemethere/upload-artifact-s3@v5
-        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && inputs.use_split_build != 'true'
+        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped'
         with:
           name: ${{ inputs.build-environment }}
           retention-days: 14
@@ -236,16 +236,6 @@ jobs:
           path: artifacts.zip
           s3-bucket: ${{ inputs.s3-bucket }}
 
-      - name: Store PyTorch Build Artifacts on S3
-        uses: seemethere/upload-artifact-s3@v5
-        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && inputs.use_split_build == 'true'
-        with:
-          name: ${{ inputs.build-environment }}-experimental-split-build
-          retention-days: 14
-          if-no-files-found: error
-          path: artifacts.zip
-          s3-bucket: ${{ inputs.s3-bucket }}
-
       - name: Upload sccache stats
         if: steps.build.outcome != 'skipped'
         uses: seemethere/upload-artifact-s3@v5

.github/workflows/_linux-test-label.yml

@@ -0,0 +1,85 @@
+name: linux-test-rg
+
+on:
+  workflow_call:
+    inputs:
+      build-environment:
+        required: true
+        type: string
+        description: Top-level label for what's being built/tested.
+      test-matrix:
+        required: true
+        type: string
+        description: JSON description of what test configs to run.
+      docker-image:
+        required: true
+        type: string
+        description: Docker image to run in.
+      sync-tag:
+        required: false
+        type: string
+        default: ""
+        description: |
+          If this is set, our linter will use this to make sure that every other
+          job with the same `sync-tag` is identical.
+      timeout-minutes:
+        required: false
+        type: number
+        default: 240
+        description: |
+          Set the maximum (in minutes) how long the workflow should take to finish
+      use-gha:
+        required: false
+        type: string
+        default: ""
+        description: If set to any value, upload to GHA. Otherwise upload to S3.
+      dashboard-tag:
+        required: false
+        type: string
+        default: ""
+      s3-bucket:
+        description: S3 bucket to download artifact
+        required: false
+        type: string
+        default: "gha-artifacts"
+      aws-role-to-assume:
+        description: role to assume for downloading artifacts
+        required: false
+        type: string
+        default: ""
+    secrets:
+      HUGGING_FACE_HUB_TOKEN:
+        required: false
+        description: |
+          HF Auth token to avoid rate limits when downloading models or datasets from hub
+
+env:
+  GIT_DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
+
+jobs:
+  test:
+    # Don't run on forked repos or empty test matrix
+    if: github.repository_owner == 'pytorch' && toJSON(fromJSON(inputs.test-matrix).include) != '[]'
+    strategy:
+      matrix: ${{ fromJSON(inputs.test-matrix) }}
+      fail-fast: false
+    runs-on: ${{ matrix.runner }}
+    timeout-minutes: ${{ matrix.mem_leak_check == 'mem_leak_check' && 600 || inputs.timeout-minutes }}
+    steps:
+      - name: Checkout PyTorch
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+
+      - name: Linux Test
+        id: linux-test
+        uses: ./.github/actions/linux-test
+        with:
+          build-environment: ${{ inputs.build-environment }}
+          test-matrix: ${{ inputs.test-matrix }}
+          docker-image: ${{ inputs.docker-image }}
+          sync-tag: ${{ inputs.sync-tag }}
+          use-gha: ${{ inputs.use-gha }}
+          dashboard-tag: ${{ inputs.dashboard-tag }}
+          s3-bucket: ${{ inputs.s3-bucket }}
+          aws-role-to-assume: ${{ inputs.aws-role-to-assume }}
+          HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/_linux-test-rg.yml

@@ -0,0 +1,86 @@
+name: linux-test-label
+
+on:
+  workflow_call:
+    inputs:
+      build-environment:
+        required: true
+        type: string
+        description: Top-level label for what's being built/tested.
+      test-matrix:
+        required: true
+        type: string
+        description: JSON description of what test configs to run.
+      docker-image:
+        required: true
+        type: string
+        description: Docker image to run in.
+      sync-tag:
+        required: false
+        type: string
+        default: ""
+        description: |
+          If this is set, our linter will use this to make sure that every other
+          job with the same `sync-tag` is identical.
+      timeout-minutes:
+        required: false
+        type: number
+        default: 240
+        description: |
+          Set the maximum (in minutes) how long the workflow should take to finish
+      use-gha:
+        required: false
+        type: string
+        default: ""
+        description: If set to any value, upload to GHA. Otherwise upload to S3.
+      dashboard-tag:
+        required: false
+        type: string
+        default: ""
+      s3-bucket:
+        description: S3 bucket to download artifact
+        required: false
+        type: string
+        default: "gha-artifacts"
+      aws-role-to-assume:
+        description: role to assume for downloading artifacts
+        required: false
+        type: string
+        default: ""
+    secrets:
+      HUGGING_FACE_HUB_TOKEN:
+        required: false
+        description: |
+          HF Auth token to avoid rate limits when downloading models or datasets from hub
+
+env:
+  GIT_DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
+
+jobs:
+  test:
+    # Don't run on forked repos or empty test matrix
+    if: github.repository_owner == 'pytorch' && toJSON(fromJSON(inputs.test-matrix).include) != '[]'
+    strategy:
+      matrix: ${{ fromJSON(inputs.test-matrix) }}
+      fail-fast: false
+    runs-on:
+      group: ${{ matrix.runner }}
+    timeout-minutes: ${{ matrix.mem_leak_check == 'mem_leak_check' && 600 || inputs.timeout-minutes }}
+    steps:
+      - name: Checkout PyTorch
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+
+      - name: Linux Test
+        id: linux-test
+        uses: ./.github/actions/linux-test
+        with:
+          build-environment: ${{ inputs.build-environment }}
+          test-matrix: ${{ inputs.test-matrix }}
+          docker-image: ${{ inputs.docker-image }}
+          sync-tag: ${{ inputs.sync-tag }}
+          use-gha: ${{ inputs.use-gha }}
+          dashboard-tag: ${{ inputs.dashboard-tag }}
+          s3-bucket: ${{ inputs.s3-bucket }}
+          aws-role-to-assume: ${{ inputs.aws-role-to-assume }}
+          HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/_runner-determinator.yml

@@ -3,23 +3,14 @@ name: Check whether the workflow owner can use ARC runners
 on:
   workflow_call:
     inputs:
-      triggering_actor:
+      user_name:
         required: true
         type: string
-        description: The triggering_actor for the workflow. Use github.triggering_actor
-      issue_owner:
-        required: true
-        type: string
-        description: The owner of the issue. Use github.event.pull_request.user.login || github.event.issue.user.login
+        description: The name of the workflow owner.
       curr_branch:
         required: true
         type: string
-        description: Current branch or tag.
-      curr_ref_type:
-        required: false
-        type: string
-        default: branch
-        description: The value of "github.ref_type", "branch" or "tag"
+        description: Current branch.
       issue_number:
         required: false
         type: string
@@ -35,240 +26,19 @@ on:
 
 jobs:
   runner-determinator:
-    runs-on: ubuntu-latest
+    runs-on: linux.4xlarge
     outputs:
       label-type: ${{ steps.set-condition.outputs.label-type }}
     env:
       GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
       ISSUE_NUMBER: ${{ inputs.issue_number }}
-      TRIGGERING_ACTOR: ${{ inputs.triggering_actor }}
-      ISSUE_OWNER: ${{ inputs.issue_owner }}
+      USERNAME: ${{ inputs.user_name }}
     steps:
-      # - name: Checkout PyTorch
-      #   uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
-      #   with:
-      #     fetch-depth: 1
-      #     submodules: true
-
-      # TODO: Remove the hardcoded step below
-      # Hardcoding below is temporary for testing ALI runners
-      # This file below should match the script found in .github/scripts/runner_determinator.py
-      - name: Hardcode runner-determinator script
-        run: |
-          cat <<EOF > runner_determinator.py
-          # flake8: noqa: G004
-
-          import logging
-          import os
-          from argparse import ArgumentParser
-          from logging import LogRecord
-          from typing import Any, Iterable
-
-          from github import Auth, Github
-          from github.Issue import Issue
-
-
-          WORKFLOW_LABEL_META = ""  # use meta runners
-          WORKFLOW_LABEL_LF = "lf."  # use runners from the linux foundation
-
-          GITHUB_OUTPUT = os.getenv("GITHUB_OUTPUT", "")
-          GH_OUTPUT_KEY_LABEL_TYPE = "label-type"
-
-
-          class ColorFormatter(logging.Formatter):
-              """Color codes the log messages based on the log level"""
-
-              COLORS = {
-                  "WARNING": "\033[33m",  # Yellow
-                  "ERROR": "\033[31m",  # Red
-                  "CRITICAL": "\033[31m",  # Red
-                  "INFO": "\033[0m",  # Reset
-                  "DEBUG": "\033[0m",  # Reset
-              }
-
-              def format(self, record: LogRecord) -> str:
-                  log_color = self.COLORS.get(record.levelname, "\033[0m")  # Default to reset
-                  record.msg = f"{log_color}{record.msg}\033[0m"
-                  return super().format(record)
-
-
-          handler = logging.StreamHandler()
-          handler.setFormatter(ColorFormatter(fmt="%(levelname)-8s: %(message)s"))
-
-          log = logging.getLogger(os.path.basename(__file__))
-          log.addHandler(handler)
-          log.setLevel(logging.INFO)
-
-
-          def set_github_output(key: str, value: str) -> None:
-              """
-              Defines outputs of the github action that invokes this script
-              """
-              if not GITHUB_OUTPUT:
-                  # See https://github.blog/changelog/2022-10-11-github-actions-deprecating-save-state-and-set-output-commands/ for deprecation notice
-                  log.warning(
-                      "No env var found for GITHUB_OUTPUT, you must be running this code locally. Falling back to the deprecated print method."
-                  )
-                  print(f"::set-output name={key}::{value}")
-                  return
-
-              with open(GITHUB_OUTPUT, "a") as f:
-                  log.info(f"Setting output: {key}='{value}'")
-                  f.write(f"{key}={value}\n")
-
-
-          def parse_args() -> Any:
-              parser = ArgumentParser("Get dynamic rollout settings")
-              parser.add_argument("--github-token", type=str, required=True, help="GitHub token")
-              parser.add_argument(
-                  "--github-issue-repo",
-                  type=str,
-                  required=False,
-                  default="pytorch/test-infra",
-                  help="GitHub repo to get the issue",
-              )
-              parser.add_argument(
-                  "--github-repo",
-                  type=str,
-                  required=True,
-                  help="GitHub repo where CI is running",
-              )
-              parser.add_argument(
-                  "--github-issue", type=int, required=True, help="GitHub issue number"
-              )
-              parser.add_argument(
-                  "--github-actor", type=str, required=True, help="GitHub triggering_actor"
-              )
-              parser.add_argument(
-                  "--github-issue-owner", type=str, required=True, help="GitHub issue owner"
-              )
-              parser.add_argument(
-                  "--github-branch", type=str, required=True, help="Current GitHub branch or tag"
-              )
-              parser.add_argument(
-                  "--github-ref-type",
-                  type=str,
-                  required=True,
-                  help="Current GitHub ref type, branch or tag",
-              )
-
-              return parser.parse_args()
-
-
-          def get_gh_client(github_token: str) -> Github:
-              auth = Auth.Token(github_token)
-              return Github(auth=auth)
-
-
-          def get_issue(gh: Github, repo: str, issue_num: int) -> Issue:
-              repo = gh.get_repo(repo)
-              return repo.get_issue(number=issue_num)
-
-
-          def get_potential_pr_author(
-              gh: Github, repo: str, username: str, ref_type: str, ref_name: str
-          ) -> str:
-              # If the trigger was a new tag added by a bot, this is a ciflow case
-              # Fetch the actual username from the original PR. The PR number is
-              # embedded in the tag name: ciflow/<name>/<pr-number>
-              if username == "pytorch-bot[bot]" and ref_type == "tag":
-                  split_tag = ref_name.split("/")
-                  if (
-                      len(split_tag) == 3
-                      and split_tag[0] == "ciflow"
-                      and split_tag[2].isnumeric()
-                  ):
-                      pr_number = split_tag[2]
-                      try:
-                          repository = gh.get_repo(repo)
-                          pull = repository.get_pull(number=int(pr_number))
-                      except Exception as e:
-                          raise Exception(  # noqa: TRY002
-                              f"issue with pull request {pr_number} from repo {repository}"
-                          ) from e
-                      return pull.user.login
-              # In all other cases, return the original input username
-              return username
-
-
-          def is_exception_branch(branch: str) -> bool:
-              return branch.split("/")[0] in {"main", "nightly", "release", "landchecks"}
-
-
-          def get_workflow_type(issue: Issue, workflow_requestors: Iterable[str]) -> str:
-              try:
-                  first_comment = issue.get_comments()[0].body.strip("\n\t ")
-
-                  if first_comment[0] == "!":
-                      log.info("LF Workflows are disabled for everyone. Using meta runners.")
-                      return WORKFLOW_LABEL_META
-                  elif first_comment[0] == "*":
-                      log.info("LF Workflows are enabled for everyone. Using LF runners.")
-                      return WORKFLOW_LABEL_LF
-                  else:
-                      all_opted_in_users = {
-                          usr_raw.strip("\n\t@ ") for usr_raw in first_comment.split()
-                      }
-                      opted_in_requestors = {
-                          usr for usr in workflow_requestors if usr in all_opted_in_users
-                      }
-                      if opted_in_requestors:
-                          log.info(
-                              f"LF Workflows are enabled for {', '.join(opted_in_requestors)}. Using LF runners."
-                          )
-                          return WORKFLOW_LABEL_LF
-                      else:
-                          log.info(
-                              f"LF Workflows are disabled for {', '.join(workflow_requestors)}. Using meta runners."
-                          )
-                          return WORKFLOW_LABEL_META
-
-              except Exception as e:
-                  log.error(
-                      f"Failed to get determine workflow type. Falling back to meta runners. Exception: {e}"
-                  )
-                  return WORKFLOW_LABEL_META
-
-
-          def main() -> None:
-              args = parse_args()
-
-              if args.github_ref_type == "branch" and is_exception_branch(args.github_branch):
-                  log.info(f"Exception branch: '{args.github_branch}', using meta runners")
-                  label_type = WORKFLOW_LABEL_META
-              else:
-                  try:
-                      gh = get_gh_client(args.github_token)
-                      # The default issue we use - https://github.com/pytorch/test-infra/issues/5132
-                      issue = get_issue(gh, args.github_issue_repo, args.github_issue)
-                      username = get_potential_pr_author(
-                          gh,
-                          args.github_repo,
-                          args.github_actor,
-                          args.github_ref_type,
-                          args.github_branch,
-                      )
-                      label_type = get_workflow_type(
-                          issue,
-                          (
-                              args.github_issue_owner,
-                              username,
-                          ),
-                      )
-                  except Exception as e:
-                      log.error(
-                          f"Failed to get issue. Falling back to meta runners. Exception: {e}"
-                      )
-                      label_type = WORKFLOW_LABEL_META
-
-              set_github_output(GH_OUTPUT_KEY_LABEL_TYPE, label_type)
-
-
-          if __name__ == "__main__":
-              main()
-          EOF
-
-          cat runner_determinator.py
+      - name: Checkout PyTorch
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        with:
+          fetch-depth: 1
+          submodules: true
 
       - name: Install dependencies
         run: python3 -m pip install urllib3==1.26.18 PyGithub==2.3.0
@@ -277,14 +47,15 @@ jobs:
         id: set-condition
         run: |
           curr_branch="${{ inputs.curr_branch }}"
-          curr_ref_type="${{ inputs.curr_ref_type }}"
           echo "Current branch is '$curr_branch'"
 
-          python3 runner_determinator.py \
+          output="$(python3 .github/scripts/get_workflow_type.py \
             --github-token "$GITHUB_TOKEN" \
             --github-issue "$ISSUE_NUMBER" \
             --github-branch "$curr_branch" \
-            --github-actor "$TRIGGERING_ACTOR" \
-            --github-issue-owner "$ISSUE_OWNER" \
-            --github-ref-type "$curr_ref_type" \
-            --github-repo "$GITHUB_REPOSITORY"
+            --github-user "$USERNAME")"
+
+          echo "Output: '${output}'"
+
+          LABEL_TYPE=$(echo "${output}" | jq -r '.label_type')
+          echo "label-type=$LABEL_TYPE" >> "$GITHUB_OUTPUT"

.github/workflows/_win-build.yml

@@ -30,12 +30,6 @@ on:
           An option JSON description of what test configs to run later on. This
           is moved here from the Linux test workflow so that we can apply filter
           logic using test-config labels earlier and skip unnecessary builds
-      runner:
-        required: false
-        type: string
-        default: "windows.4xlarge.nonephemeral"
-        description: |
-          Label of the runner this job should run on.
 
     outputs:
       test-matrix:
@@ -49,7 +43,7 @@ jobs:
   build:
     # Don't run on forked repos.
     if: github.repository_owner == 'pytorch'
-    runs-on: ${{ inputs.runner }}
+    runs-on: [self-hosted, windows.4xlarge.nonephemeral]
     timeout-minutes: 240
     outputs:
       test-matrix: ${{ steps.filter.outputs.test-matrix }}

.github/workflows/build-triton-wheel.yml

@@ -14,7 +14,6 @@ on:
       - .github/ci_commit_pins/triton.txt
       - .ci/docker/ci_commit_pins/triton.txt
       - .ci/docker/ci_commit_pins/triton-rocm.txt
-      - .ci/docker/ci_commit_pins/triton-xpu.txt
   pull_request:
     paths:
       - .github/workflows/build-triton-wheel.yml
@@ -22,7 +21,6 @@ on:
       - .github/ci_commit_pins/triton.txt
       - .ci/docker/ci_commit_pins/triton.txt
       - .ci/docker/ci_commit_pins/triton-rocm.txt
-      - .ci/docker/ci_commit_pins/triton-xpu.txt
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name == 'workflow_dispatch' }}
@@ -36,7 +34,7 @@ jobs:
       fail-fast: false
       matrix:
         py_vers: [ "3.8", "3.9", "3.10", "3.11", "3.12" ]
-        device: ["cuda", "rocm", "xpu"]
+        device: ["cuda", "rocm"]
         include:
           - device: "rocm"
             rocm_version: "6.1"
@@ -104,6 +102,11 @@ jobs:
             ;;
           esac
 
+          BUILD_ROCM=""
+          if [[ "$BUILD_DEVICE" == "rocm" ]]; then
+            BUILD_ROCM="--build-rocm"
+          fi
+
           RELEASE=""
           if [[ "${IS_RELEASE_TAG}" == true ]]; then
             RELEASE="--release"
@@ -111,13 +114,7 @@ jobs:
 
           docker exec -t "${container_name}" yum install -y zlib-devel zip
           docker exec -t "${container_name}" "${PYTHON_EXECUTABLE}"  -m pip install -U setuptools==67.4.0
-          # Triton xpu build use GCC11
-          if [[ "${BUILD_DEVICE}" == xpu ]]; then
-            docker exec -t "${container_name}" yum install -y devtoolset-11-gcc-c++
-            docker exec -t "${container_name}" bash -c "source /opt/rh/devtoolset-11/enable && ${PYTHON_EXECUTABLE} /pytorch/.github/scripts/build_triton_wheel.py --device=$BUILD_DEVICE $RELEASE"
-          else
-            docker exec -t "${container_name}" bash -c "${PYTHON_EXECUTABLE} /pytorch/.github/scripts/build_triton_wheel.py --device=$BUILD_DEVICE $RELEASE"
-          fi
+          docker exec -t "${container_name}" "${PYTHON_EXECUTABLE}" /pytorch/.github/scripts/build_triton_wheel.py $BUILD_ROCM $RELEASE
           docker exec -t "${container_name}" chown -R 1000.1000 /artifacts
 
       - uses: actions/upload-artifact@v3

.github/workflows/create_release.yml

@@ -5,11 +5,6 @@ on:
     branches:
       - main
       - release/*
-    tags:
-      # Final Release tags look like: v1.11.0
-      - v[0-9]+.[0-9]+.[0-9]+
-      # Release candidate tags look like: v1.11.0-rc1
-      - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
   release:
     types: [published]
   pull_request:
@@ -23,8 +18,6 @@ jobs:
     # https://github.com/softprops/action-gh-release?tab=readme-ov-file#permissions
     permissions:
       contents: write
-    outputs:
-      pt_release_name: ${{ steps.release_name.outputs.pt_release_name }}
     steps:
       - uses: malfet/checkout@silent-checkout
         with:
@@ -56,44 +49,11 @@ jobs:
             # Create archive
             tar -czf "$PT_RELEASE_FILE" "$PT_RELEASE_NAME"
             echo "Created source archive $PT_RELEASE_FILE with content: $(ls -a "$PT_RELEASE_NAME")"
-      - name: Upload source distribution for release
+      - name: Upload source distribution
         if: ${{ github.event_name == 'release' }}
         uses: softprops/action-gh-release@v1
         with:
           files: ${{env.PT_RELEASE_FILE}}
-      - name: Upload source distribution to GHA artifacts for release tags
-        if: ${{ github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v') && contains(github.ref, 'rc') }}
-        uses: actions/upload-artifact@v2
-        with:
-          name: ${{ env.PT_RELEASE_FILE }}
-          path: ${{ env.PT_RELEASE_FILE }}
-      - name: Set output
-        id: release_name
-        run: echo "::set-output name=pt_release_name::${{ env.PT_RELEASE_NAME }}.tar.gz"
-
-  upload_source_code_to_s3:
-    if: ${{ github.repository == 'pytorch/pytorch' && github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v') && contains(github.ref, 'rc') }}
-    runs-on: linux.2xlarge
-    environment: sourcecode-upload
-    name: Upload source code to S3 for release tags
-    permissions:
-      id-token: write
-    needs: release
-    steps:
-      - uses: actions/download-artifact@v2
-        with:
-          name: ${{ needs.release.outputs.pt_release_name }}
-      - name: Configure AWS credentials(PyTorch account)
-        uses: aws-actions/configure-aws-credentials@v3
-        with:
-          role-to-assume: arn:aws:iam::749337293305:role/gha_pytorch_source_code_upload_role
-          aws-region: us-east-1
-      - uses: seemethere/upload-artifact-s3@v5
-        with:
-          s3-bucket: pytorch
-          s3-prefix: source_code/test
-          if-no-files-found: warn
-          path: ${{ needs.release.outputs.pt_release_name }}
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name }}

.github/workflows/generated-linux-aarch64-binary-manywheel-nightly.yml

@@ -54,7 +54,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_8-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_8-cpu-aarch64-test:  # Testing
@@ -162,7 +162,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_9-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cpu-aarch64-test:  # Testing
@@ -270,7 +270,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_10-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_10-cpu-aarch64-test:  # Testing
@@ -378,7 +378,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_11-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_11-cpu-aarch64-test:  # Testing
@@ -486,7 +486,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_12-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_12-cpu-aarch64-test:  # Testing

.github/workflows/generated-linux-binary-manywheel-main.yml

@@ -48,7 +48,7 @@ jobs:
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda11_8
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu11==11.8.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu11==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu11==11.11.3.6; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu11==10.9.0.58; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu11==10.3.0.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu11==11.4.1.48; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu11==11.7.5.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu11==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu11==11.8.86; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu11==11.8.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu11==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu11==11.11.3.6; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu11==10.9.0.58; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu11==10.3.0.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu11==11.4.1.48; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu11==11.7.5.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu11==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu11==11.8.86; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_8-cuda11_8-test:  # Testing
@@ -72,48 +72,6 @@ jobs:
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
 
-  manywheel-py3_8-cuda11_8-split-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu118
-      GPU_ARCH_VERSION: 11.8
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.8"
-      build_name: manywheel-py3_8-cuda11_8-split
-      build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu11==11.8.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu11==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu11==11.11.3.6; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu11==10.9.0.58; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu11==10.3.0.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu11==11.4.1.48; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu11==11.7.5.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu11==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu11==11.8.86; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_8-cuda11_8-split-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs: manywheel-py3_8-cuda11_8-split-build
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu118
-      GPU_ARCH_VERSION: 11.8
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.8"
-      build_name: manywheel-py3_8-cuda11_8-split
-      build_environment: linux-binary-manywheel
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-
   manywheel-py3_8-cuda12_1-build:
     if: ${{ github.repository_owner == 'pytorch' }}
     uses: ./.github/workflows/_binary-build-linux.yml
@@ -130,7 +88,7 @@ jobs:
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda12_1
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_8-cuda12_1-test:  # Testing
@@ -154,48 +112,6 @@ jobs:
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
 
-  manywheel-py3_8-cuda12_1-split-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu121
-      GPU_ARCH_VERSION: 12.1
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.8"
-      build_name: manywheel-py3_8-cuda12_1-split
-      build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_8-cuda12_1-split-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs: manywheel-py3_8-cuda12_1-split-build
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu121
-      GPU_ARCH_VERSION: 12.1
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.8"
-      build_name: manywheel-py3_8-cuda12_1-split
-      build_environment: linux-binary-manywheel
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-
   manywheel-py3_8-cuda12_4-build:
     if: ${{ github.repository_owner == 'pytorch' }}
     uses: ./.github/workflows/_binary-build-linux.yml
@@ -212,7 +128,7 @@ jobs:
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda12_4
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.2.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.0.44; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.119; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.0.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.0.142; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.2.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.0.44; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.119; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.0.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.0.142; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_8-cuda12_4-test:  # Testing
@@ -235,45 +151,3 @@ jobs:
       runs_on: linux.4xlarge.nvidia.gpu
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
-
-  manywheel-py3_8-cuda12_4-split-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu124
-      GPU_ARCH_VERSION: 12.4
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.8"
-      build_name: manywheel-py3_8-cuda12_4-split
-      build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.2.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.0.44; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.119; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.0.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.0.142; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_8-cuda12_4-split-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs: manywheel-py3_8-cuda12_4-split-build
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu124
-      GPU_ARCH_VERSION: 12.4
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.8"
-      build_name: manywheel-py3_8-cuda12_4-split
-      build_environment: linux-binary-manywheel
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/generated-linux-s390x-binary-manywheel-nightly.yml

@@ -54,7 +54,7 @@ jobs:
       ALPINE_IMAGE: "docker.io/s390x/alpine"
       build_name: manywheel-py3_8-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_8-cpu-s390x-test:  # Testing
@@ -117,7 +117,7 @@ jobs:
       ALPINE_IMAGE: "docker.io/s390x/alpine"
       build_name: manywheel-py3_9-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cpu-s390x-test:  # Testing
@@ -180,7 +180,7 @@ jobs:
       ALPINE_IMAGE: "docker.io/s390x/alpine"
       build_name: manywheel-py3_10-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_10-cpu-s390x-test:  # Testing
@@ -243,7 +243,7 @@ jobs:
       ALPINE_IMAGE: "docker.io/s390x/alpine"
       build_name: manywheel-py3_11-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_11-cpu-s390x-test:  # Testing
@@ -306,7 +306,7 @@ jobs:
       ALPINE_IMAGE: "docker.io/s390x/alpine"
       build_name: manywheel-py3_12-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_12-cpu-s390x-test:  # Testing

.github/workflows/generated-macos-arm64-binary-wheel-nightly.yml

@@ -46,7 +46,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.8"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
       # For sccache access (only on non-forked PRs)
       AWS_ACCESS_KEY_ID: ${{ secrets.MACOS_SCCACHE_S3_ACCESS_KEY_ID }}
       AWS_SECRET_ACCESS_KEY: ${{ secrets.MACOS_SCCACHE_S3_SECRET_ACCESS_KEY }}
@@ -165,7 +165,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.9"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
       # For sccache access (only on non-forked PRs)
       AWS_ACCESS_KEY_ID: ${{ secrets.MACOS_SCCACHE_S3_ACCESS_KEY_ID }}
       AWS_SECRET_ACCESS_KEY: ${{ secrets.MACOS_SCCACHE_S3_SECRET_ACCESS_KEY }}
@@ -284,7 +284,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.10"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
       # For sccache access (only on non-forked PRs)
       AWS_ACCESS_KEY_ID: ${{ secrets.MACOS_SCCACHE_S3_ACCESS_KEY_ID }}
       AWS_SECRET_ACCESS_KEY: ${{ secrets.MACOS_SCCACHE_S3_SECRET_ACCESS_KEY }}
@@ -403,7 +403,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.11"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
       # For sccache access (only on non-forked PRs)
       AWS_ACCESS_KEY_ID: ${{ secrets.MACOS_SCCACHE_S3_ACCESS_KEY_ID }}
       AWS_SECRET_ACCESS_KEY: ${{ secrets.MACOS_SCCACHE_S3_SECRET_ACCESS_KEY }}
@@ -522,7 +522,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.12"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
       # For sccache access (only on non-forked PRs)
       AWS_ACCESS_KEY_ID: ${{ secrets.MACOS_SCCACHE_S3_ACCESS_KEY_ID }}
       AWS_SECRET_ACCESS_KEY: ${{ secrets.MACOS_SCCACHE_S3_SECRET_ACCESS_KEY }}

.github/workflows/generated-windows-binary-wheel-nightly.yml

@@ -46,7 +46,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.8"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -290,7 +290,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.8"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -536,7 +536,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.8"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -782,7 +782,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.8"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -1027,7 +1027,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.9"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -1271,7 +1271,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.9"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -1517,7 +1517,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.9"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -1763,7 +1763,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.9"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -2008,7 +2008,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.10"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -2252,7 +2252,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.10"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -2498,7 +2498,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.10"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -2744,7 +2744,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.10"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -2989,7 +2989,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.11"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -3233,7 +3233,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.11"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -3479,7 +3479,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.11"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -3725,7 +3725,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.11"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -3970,7 +3970,7 @@ jobs:
       GPU_ARCH_TYPE: cpu
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.12"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -4214,7 +4214,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.12"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -4460,7 +4460,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.12"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash
@@ -4706,7 +4706,7 @@ jobs:
       GPU_ARCH_TYPE: cuda
       SKIP_ALL_TESTS: 1
       DESIRED_PYTHON: "3.12"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
     steps:
       - name: Display EC2 information
         shell: bash

.github/workflows/inductor.yml

@@ -24,8 +24,7 @@ jobs:
       docker-image-name: pytorch-linux-focal-rocm-n-py3
       test-matrix: |
         { include: [
-          { config: "inductor", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.2" },
-          { config: "inductor", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.2" },
+          { config: "inductor", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.2" },
         ]}
 
   linux-focal-rocm6_1-py3_8-inductor-test:
@@ -172,21 +171,11 @@ jobs:
           { config: "cpu_inductor_timm_freezing", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
           { config: "cpu_inductor_torchbench_freezing", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
           { config: "cpu_inductor_torchbench_freezing", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_inductor_huggingface_amp_freezing", shard: 1, num_shards: 1, runner: "linux.16xlarge.spr" },
-          { config: "cpu_inductor_timm_amp_freezing", shard: 1, num_shards: 2, runner: "linux.16xlarge.spr" },
-          { config: "cpu_inductor_timm_amp_freezing", shard: 2, num_shards: 2, runner: "linux.16xlarge.spr" },
-          { config: "cpu_inductor_torchbench_amp_freezing", shard: 1, num_shards: 2, runner: "linux.16xlarge.spr" },
-          { config: "cpu_inductor_torchbench_amp_freezing", shard: 2, num_shards: 2, runner: "linux.16xlarge.spr" },
           { config: "dynamic_cpu_inductor_huggingface", shard: 1, num_shards: 1, runner: "linux.12xlarge" },
           { config: "dynamic_cpu_inductor_timm", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
           { config: "dynamic_cpu_inductor_timm", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
           { config: "dynamic_cpu_inductor_torchbench", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
           { config: "dynamic_cpu_inductor_torchbench", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_huggingface_freezing", shard: 1, num_shards: 1, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_timm_freezing", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_timm_freezing", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_torchbench_freezing", shard: 1, num_shards: 2, runner: "linux.12xlarge" },
-          { config: "cpu_aot_inductor_torchbench_freezing", shard: 2, num_shards: 2, runner: "linux.12xlarge" },
           { config: "inductor_torchbench_cpu_smoketest_perf", shard: 1, num_shards: 1, runner: "linux.24xl.spr-metal" },
         ]}
     secrets:

.github/workflows/lint.yml

@@ -19,10 +19,10 @@ jobs:
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       timeout: 120
-      runner: linux.2xlarge
+      runner: lf.linux.2xlarge
       docker-image: pytorch-linux-jammy-cuda11.8-cudnn9-py3.9-linter
       # NB: A shallow checkout won't work here because calculate-docker-image requires a full checkout
-      # to run git rev-parse HEAD~:.ci/docker when a new image is needed
+      # to run git rev-parse HEAD~:.ci/docker when a new image is needed.
       fetch-depth: 0
       submodules: true
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
@@ -35,7 +35,7 @@ jobs:
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       timeout: 120
-      runner: linux.2xlarge
+      runner: lf.linux.2xlarge
       docker-image: pytorch-linux-jammy-cuda11.8-cudnn9-py3.9-linter
       # NB: A shallow checkout won't work here because calculate-docker-image requires a full checkout
       # to run git rev-parse HEAD~:.ci/docker when a new image is needed
@@ -49,7 +49,7 @@ jobs:
   quick-checks:
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
-      runner: linux.2xlarge
+      runner: lf.linux.2xlarge
       docker-image: pytorch-linux-focal-linter
       fetch-depth: 0
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}

.github/workflows/llm_td_retrieval.yml

@@ -36,24 +36,33 @@ jobs:
           ref: v0.0.2
           path: llm-target-determinator
 
-      - name: Setup miniconda
-        uses: pytorch/test-infra/.github/actions/setup-miniconda@main
+      - name: Setup Conda
+        uses: conda-incubator/setup-miniconda@v2.1.1
         with:
-          python-version: "3.9"
+          miniconda-version: "py39_4.12.0"
+          python-version: 3.9
 
-      - name: Install requirements
+      - name: Install Requirements
         shell: bash -l {0}
         run: |
           set -euxo pipefail
-          ${CONDA_RUN} pip install -r llm-target-determinator/requirements.txt
-          cd "${GITHUB_WORKSPACE}/codellama"
-          ${CONDA_RUN} pip install -e .
+          conda create \
+            --yes \
+            --quiet \
+            --name "tdenv" \
+            "python=3.9"
+          conda activate tdenv
+          cd "${GITHUB_WORKSPACE}/llm-target-determinator"
+          pip install -r requirements.txt
+          cd ../codellama
+          pip install -e .
 
       - name: Fetch CodeLlama Checkpoint
         shell: bash -l {0}
         run: |
           set -euxo pipefail
-          cd "${GITHUB_WORKSPACE}/codellama"
+          conda activate tdenv
+          cd codellama/
           mkdir "CodeLlama-7b-Python"
           aws s3 cp "s3://target-determinator-assets/CodeLlama-7b-Python" "CodeLlama-7b-Python" --recursive --no-progress
 
@@ -66,7 +75,7 @@ jobs:
           shell: bash
           command: |
             set -euxo pipefail
-            ${CONDA_RUN} python -m pip install awscli==1.29.40
+            python3 -m pip install awscli==1.29.40
             cd "${GITHUB_WORKSPACE}"/llm-target-determinator/assets
             aws s3 cp "s3://target-determinator-assets/indexes/latest" . --recursive
 
@@ -79,8 +88,9 @@ jobs:
         shell: bash -l {0}
         run: |
           set -euxo pipefail
+          conda activate tdenv
           cd "${GITHUB_WORKSPACE}"/llm-target-determinator
-          ${CONDA_RUN} torchrun \
+          torchrun \
             --standalone \
             --nnodes=1 \
             --nproc-per-node=1 \

.github/workflows/periodic.yml

@@ -294,53 +294,3 @@ jobs:
       build-environment: linux-focal-rocm6.1-py3.8
       docker-image: ${{ needs.linux-focal-rocm6_1-py3_8-build.outputs.docker-image }}
       test-matrix: ${{ needs.linux-focal-rocm6_1-py3_8-build.outputs.test-matrix }}
-
-  linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build:
-    name: linux-focal-cuda12.1-py3.10-gcc9-experimental-split-build
-    uses: ./.github/workflows/_linux-build-label.yml
-    with:
-      use_split_build: true
-      build-environment: linux-focal-cuda12.1-py3.10-gcc9
-      docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9
-      test-matrix: |
-        { include: [
-          { config: "nogpu_AVX512", shard: 1, num_shards: 1, runner: "linux.2xlarge" },
-          { config: "nogpu_NO_AVX2", shard: 1, num_shards: 1, runner: "linux.2xlarge" },
-          { config: "jit_legacy", shard: 1, num_shards: 1, runner: "linux.4xlarge.nvidia.gpu" },
-        ]}
-  linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build-test:
-    name: linux-focal-cuda12.1-py3.10-gcc9-experimental-split-build
-    uses: ./.github/workflows/_linux-test.yml
-    needs:
-      - linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build
-      - target-determination
-    with:
-      build-environment: linux-focal-cuda12.1-py3.10-gcc9-experimental-split-build
-      docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build.outputs.test-matrix }}
-
-
-  linux-focal-cuda11_8-py3_9-gcc9-experimental-split-build:
-    name: linux-focal-cuda11.8-py3.9-gcc9-experimental-split-build
-    uses: ./.github/workflows/_linux-build-label.yml
-    with:
-      use_split_build: true
-      build-environment: linux-focal-cuda11.8-py3.9-gcc9
-      docker-image-name: pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9
-      cuda-arch-list: 8.6
-      test-matrix: |
-        { include: [
-          { config: "multigpu", shard: 1, num_shards: 1, runner: "linux.g5.12xlarge.nvidia.gpu" },
-        ]}
-      build-with-debug: false
-
-  linux-focal-cuda11_8-py3_9-gcc9-experimental-split-build-test:
-    name: linux-focal-cuda11.8-py3.9-gcc9-experimental-split-build-test
-    uses: ./.github/workflows/_linux-test.yml
-    needs:
-      - linux-focal-cuda11_8-py3_9-gcc9-experimental-split-build
-      - target-determination
-    with:
-      build-environment: linux-focal-cuda11.8-py3.9-gcc9-experimental-split-build
-      docker-image: ${{ needs.linux-focal-cuda11_8-py3_9-gcc9-experimental-split-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-focal-cuda11_8-py3_9-gcc9-experimental-split-build.outputs.test-matrix }}

.github/workflows/pull.yml

@@ -35,33 +35,22 @@ jobs:
       id-token: write
       contents: read
 
-  get-label-type:
-    name: get-label-type
-    uses: ./.github/workflows/_runner-determinator.yml
-    with:
-      triggering_actor: ${{ github.triggering_actor }}
-      issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
-      curr_branch: ${{ github.head_ref || github.ref_name }}
-
   linux-jammy-py3_8-gcc11-build:
     name: linux-jammy-py3.8-gcc11
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-py3.8-gcc11
       docker-image-name: pytorch-linux-jammy-py3.8-gcc11
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 2, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 3, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 4, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "docs_test", shard: 1, num_shards: 1,  runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "jit_legacy", shard: 1, num_shards: 1, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "backwards_compat", shard: 1, num_shards: 1, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "distributed", shard: 1, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "distributed", shard: 2, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
+          { config: "default", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "default", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "default", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "docs_test", shard: 1, num_shards: 1,  runner: "linux.2xlarge" },
+          { config: "jit_legacy", shard: 1, num_shards: 1, runner: "linux.2xlarge" },
+          { config: "backwards_compat", shard: 1, num_shards: 1, runner: "linux.2xlarge" },
+          { config: "distributed", shard: 1, num_shards: 2, runner: "linux.2xlarge" },
+          { config: "distributed", shard: 2, num_shards: 2, runner: "linux.2xlarge" },
         ]}
 
   linux-jammy-py3_8-gcc11-test:
@@ -86,9 +75,7 @@ jobs:
   linux-jammy-py3_8-gcc11-no-ops:
     name: linux-jammy-py3.8-gcc11-no-ops
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-py3.8-gcc11-no-ops
       docker-image-name: pytorch-linux-jammy-py3.8-gcc11
       test-matrix: |
@@ -99,9 +86,7 @@ jobs:
   linux-jammy-py3_8-gcc11-pch:
     name: linux-jammy-py3.8-gcc11-pch
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-py3.8-gcc11-pch
       docker-image-name: pytorch-linux-jammy-py3.8-gcc11
       test-matrix: |
@@ -113,19 +98,17 @@ jobs:
   linux-jammy-py3_10-clang15-asan-build:
     name: linux-jammy-py3.10-clang15-asan
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-py3.10-clang15-asan
       docker-image-name: pytorch-linux-jammy-py3-clang15-asan
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 6, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
-          { config: "default", shard: 2, num_shards: 6, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
-          { config: "default", shard: 3, num_shards: 6, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
-          { config: "default", shard: 4, num_shards: 6, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
-          { config: "default", shard: 5, num_shards: 6, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
-          { config: "default", shard: 6, num_shards: 6, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
+          { config: "default", shard: 1, num_shards: 6, runner: "linux.4xlarge" },
+          { config: "default", shard: 2, num_shards: 6, runner: "linux.4xlarge" },
+          { config: "default", shard: 3, num_shards: 6, runner: "linux.4xlarge" },
+          { config: "default", shard: 4, num_shards: 6, runner: "linux.4xlarge" },
+          { config: "default", shard: 5, num_shards: 6, runner: "linux.4xlarge" },
+          { config: "default", shard: 6, num_shards: 6, runner: "linux.4xlarge" },
         ]}
       sync-tag: asan-build
 
@@ -145,15 +128,13 @@ jobs:
   linux-focal-py3_8-clang10-onnx-build:
     name: linux-focal-py3.8-clang10-onnx
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-py3.8-clang10-onnx
       docker-image-name: pytorch-linux-focal-py3-clang10-onnx
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 2, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
+          { config: "default", shard: 1, num_shards: 2, runner: "linux.2xlarge" },
+          { config: "default", shard: 2, num_shards: 2, runner: "linux.2xlarge" },
         ]}
 
   linux-focal-py3_8-clang10-onnx-test:
@@ -170,22 +151,19 @@ jobs:
   linux-focal-py3_8-clang10-build:
     name: linux-focal-py3.8-clang10
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-py3.8-clang10
       docker-image-name: pytorch-linux-focal-py3.8-clang10
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 2, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 3, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 4, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "crossref", shard: 1, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "crossref", shard: 2, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
+          { config: "default", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "default", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "default", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "crossref", shard: 1, num_shards: 2, runner: "linux.2xlarge" },
+          { config: "crossref", shard: 2, num_shards: 2, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
         ]}
   linux-focal-py3_8-clang10-test:
     name: linux-focal-py3.8-clang10
@@ -201,24 +179,22 @@ jobs:
   linux-focal-py3_11-clang10-build:
     name: linux-focal-py3.11-clang10
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-py3.11-clang10
       docker-image-name: pytorch-linux-focal-py3.11-clang10
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 2, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 3, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 4, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "crossref", shard: 1, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "crossref", shard: 2, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
+          { config: "default", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "default", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "default", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "crossref", shard: 1, num_shards: 2, runner: "linux.2xlarge" },
+          { config: "crossref", shard: 2, num_shards: 2, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
         ]}
 
+
   linux-focal-py3_11-clang10-test:
     name: linux-focal-py3.11-clang10
     uses: ./.github/workflows/_linux-test.yml
@@ -233,20 +209,17 @@ jobs:
   linux-focal-py3_12-clang10-build:
     name: linux-focal-py3.12-clang10
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-py3.12-clang10
       docker-image-name: pytorch-linux-focal-py3.12-clang10
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 2, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 3, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "default", shard: 4, num_shards: 4, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
-          { config: "dynamo", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
+          { config: "default", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "default", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "default", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
+          { config: "dynamo", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
         ]}
 
   linux-focal-py3_12-clang10-test:
@@ -262,16 +235,14 @@ jobs:
   linux-focal-cuda11_8-py3_10-gcc9-build:
     name: linux-focal-cuda11.8-py3.10-gcc9
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-cuda11.8-py3.10-gcc9
       docker-image-name: pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9
       test-matrix: |
         { include: [
-          { config: "distributed", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
-          { config: "distributed", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
-          { config: "distributed", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
+          { config: "distributed", shard: 1, num_shards: 3, runner: "linux.8xlarge.nvidia.gpu" },
+          { config: "distributed", shard: 2, num_shards: 3, runner: "linux.8xlarge.nvidia.gpu" },
+          { config: "distributed", shard: 3, num_shards: 3, runner: "linux.8xlarge.nvidia.gpu" },
         ]}
 
   linux-focal-cuda11_8-py3_10-gcc9-test:
@@ -289,18 +260,16 @@ jobs:
   linux-focal-cuda12_1-py3_10-gcc9-build:
     name: linux-focal-cuda12.1-py3.10-gcc9
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-cuda12.1-py3.10-gcc9
       docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 2, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 3, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 4, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 5, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 1, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 2, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 3, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 4, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 5, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
         ]}
 
   linux-focal-cuda12_1-py3_10-gcc9-test:
@@ -318,9 +287,7 @@ jobs:
   linux-jammy-py3-clang12-mobile-build:
     name: linux-jammy-py3-clang12-mobile-build
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-py3-clang12-mobile-build
       docker-image-name: pytorch-linux-jammy-py3-clang15-asan
       build-generates-artifacts: false
@@ -332,9 +299,7 @@ jobs:
   linux-jammy-cuda-11_8-cudnn9-py3_8-clang12-build:
     name: linux-jammy-cuda11.8-cudnn9-py3.8-clang12
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-cuda11.8-cudnn9-py3.8-clang12
       docker-image-name: pytorch-linux-jammy-cuda11.8-cudnn9-py3.8-clang12
       test-matrix: |
@@ -345,9 +310,7 @@ jobs:
   linux-focal-py3-clang9-mobile-custom-build-static:
     name: linux-focal-py3-clang9-mobile-custom-build-static
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-py3-clang9-mobile-custom-build-static
       docker-image-name: pytorch-linux-focal-py3-clang9-android-ndk-r21e
       build-generates-artifacts: false
@@ -359,14 +322,12 @@ jobs:
   linux-focal-py3_8-clang9-xla-build:
     name: linux-focal-py3_8-clang9-xla
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-py3.8-clang9-xla
       docker-image-name: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/xla_base:v1.1-lite
       test-matrix: |
         { include: [
-          { config: "xla", shard: 1, num_shards: 1, runner: "${{ needs.get-label-type.outputs.label-type }}linux.12xlarge" },
+          { config: "xla", shard: 1, num_shards: 1, runner: "linux.12xlarge" },
         ]}
 
   linux-focal-py3_8-clang9-xla-test:
@@ -383,59 +344,51 @@ jobs:
     if: github.event_name == 'pull_request'
     name: win-vs2019-cpu-py3
     uses: ./.github/workflows/_win-build.yml
-    needs: get-label-type
     with:
       build-environment: win-vs2019-cpu-py3
       cuda-version: cpu
       sync-tag: win-cpu-build
-      runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral"
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral" },
-          { config: "default", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral" },
-          { config: "default", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral" },
+          { config: "default", shard: 1, num_shards: 3, runner: "windows.4xlarge.nonephemeral" },
+          { config: "default", shard: 2, num_shards: 3, runner: "windows.4xlarge.nonephemeral" },
+          { config: "default", shard: 3, num_shards: 3, runner: "windows.4xlarge.nonephemeral" },
         ]}
 
   linux-focal-cpu-py3_10-gcc9-bazel-test:
     name: linux-focal-cpu-py3.10-gcc9-bazel-test
     uses: ./.github/workflows/_bazel-build-test.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.large"
       build-environment: linux-focal-cuda12.1-py3.10-gcc9-bazel-test
       docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9
       cuda-version: cpu
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 1, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
+          { config: "default", shard: 1, num_shards: 1, runner: "linux.4xlarge" },
         ]}
 
   linux-focal-cuda12_1-py3_10-gcc9-bazel-test:
     name: linux-focal-cuda12.1-py3.10-gcc9-bazel-test
     uses: ./.github/workflows/_bazel-build-test.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.large"
       build-environment: linux-focal-cuda12.1-py3.10-gcc9-bazel-test
       docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9
       cuda-version: "12.1"
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 1, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 1, num_shards: 1, runner: "linux.4xlarge.nvidia.gpu" },
         ]}
 
   linux-focal-cuda12_4-py3_10-gcc9-bazel-test:
     name: linux-focal-cuda12.4-py3.10-gcc9-bazel-test
     uses: ./.github/workflows/_bazel-build-test.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.large"
       build-environment: linux-focal-cuda12.4-py3.10-gcc9-bazel-test
       docker-image-name: pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9
       cuda-version: "12.4"
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 1, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 1, num_shards: 1, runner: "linux.4xlarge.nvidia.gpu" },
         ]}
 
   linux-focal-py3-clang9-android-ndk-r21e-gradle-custom-build-single:
@@ -463,9 +416,7 @@ jobs:
   linux-jammy-py3_8-gcc11-mobile-lightweight-dispatch-build:
     name: linux-jammy-py3.8-gcc11-mobile-lightweight-dispatch-build
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-py3.8-gcc111-mobile-lightweight-dispatch-build
       docker-image-name: pytorch-linux-jammy-py3.8-gcc11
       build-generates-artifacts: false
@@ -479,9 +430,7 @@ jobs:
     if: github.event_name == 'pull_request'
     name: linux-focal-rocm6.1-py3.8
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-rocm6.1-py3.8
       docker-image-name: pytorch-linux-focal-rocm-n-py3
       sync-tag: rocm-build
@@ -495,19 +444,17 @@ jobs:
   linux-focal-cuda12_1-py3_10-gcc9-sm86-build:
     name: linux-focal-cuda12.1-py3.10-gcc9-sm86
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-cuda12.1-py3.10-gcc9-sm86
       docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9
       cuda-arch-list: 8.6
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g5.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 2, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g5.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 3, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g5.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 4, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g5.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 5, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 1, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 2, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 3, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 4, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 5, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
         ]}
 
   linux-focal-cuda12_1-py3_10-gcc9-sm86-test:
@@ -524,14 +471,12 @@ jobs:
   linux-jammy-py3-clang12-executorch-build:
     name: linux-jammy-py3-clang12-executorch
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-py3-clang12-executorch
       docker-image-name: pytorch-linux-jammy-py3-clang12-executorch
       test-matrix: |
         { include: [
-          { config: "executorch", shard: 1, num_shards: 1, runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge" },
+          { config: "executorch", shard: 1, num_shards: 1, runner: "linux.2xlarge" },
         ]}
 
   linux-jammy-py3-clang12-executorch-test:
@@ -546,22 +491,20 @@ jobs:
   linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build:
     name: linux-focal-cuda12.1-py3.10-gcc9-experimental-split-build
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       use_split_build: true
       build-environment: linux-focal-cuda12.1-py3.10-gcc9
       docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 2, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 3, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 4, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 5, num_shards: 5, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 1, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 2, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 3, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 4, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 5, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
         ]}
 
-  linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build-test:
+  linux-focal-cuda12_4-py3_10-gcc9-experimental-split-build-test:
     name: linux-focal-cuda12.1-py3.10-gcc9-experimental-split-build
     uses: ./.github/workflows/_linux-test.yml
     needs:
@@ -572,29 +515,3 @@ jobs:
       build-environment: linux-focal-cuda12.1-py3.10-gcc9-experimental-split-build
       docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build.outputs.docker-image }}
       test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-experimental-split-build.outputs.test-matrix }}
-
-  linux-focal-py3_12-clang10-experimental-split-build:
-    name: linux-focal-py3.12-clang10-experimental-split-build
-    uses: ./.github/workflows/_linux-build-label.yml
-    with:
-      use_split_build: True
-      build-environment: linux-focal-py3.12-clang10
-      docker-image-name: pytorch-linux-focal-py3.12-clang10
-      test-matrix: |
-        { include: [
-          { config: "default", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
-          { config: "default", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
-          { config: "default", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
-          { config: "dynamo", shard: 1, num_shards: 3, runner: "linux.2xlarge" },
-          { config: "dynamo", shard: 2, num_shards: 3, runner: "linux.2xlarge" },
-          { config: "dynamo", shard: 3, num_shards: 3, runner: "linux.2xlarge" },
-        ]}
-  linux-focal-py3_12-clang10-experimental-split-build-test:
-    name: linux-focal-py3.12-clang10-experimental-split-build
-    uses: ./.github/workflows/_linux-test.yml
-    needs: linux-focal-py3_12-clang10-experimental-split-build
-    with:
-      build-environment: linux-focal-py3.12-clang10-experimental-split-build
-      docker-image: ${{ needs.linux-focal-py3_12-clang10-experimental-split-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-focal-py3_12-clang10-experimental-split-build.outputs.test-matrix }}
-      timeout-minutes: 600

.github/workflows/slow.yml

@@ -36,15 +36,6 @@ jobs:
       id-token: write
       contents: read
 
-  get-label-type:
-    name: get-label-type
-    uses: ./.github/workflows/_runner-determinator.yml
-    with:
-      triggering_actor: ${{ github.triggering_actor }}
-      issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
-      curr_branch: ${{ github.head_ref || github.ref_name }}
-      curr_ref_type: ${{ github.ref_type }}
-
   linux-focal-cuda12_1-py3-gcc9-slow-gradcheck-build:
     name: linux-focal-cuda12.1-py3-gcc9-slow-gradcheck
     uses: ./.github/workflows/_linux-build.yml
@@ -150,16 +141,14 @@ jobs:
   linux-jammy-py3_10-clang15-asan-build:
     name: linux-jammy-py3.10-clang15-asan
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-jammy-py3.10-clang15-asan
       docker-image-name: pytorch-linux-jammy-py3-clang15-asan
       test-matrix: |
         { include: [
-          { config: "slow", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
-          { config: "slow", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
-          { config: "slow", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.4xlarge" },
+          { config: "slow", shard: 1, num_shards: 3, runner: "linux.4xlarge" },
+          { config: "slow", shard: 2, num_shards: 3, runner: "linux.4xlarge" },
+          { config: "slow", shard: 3, num_shards: 3, runner: "linux.4xlarge" },
         ]}
       sync-tag: asan-build
 

.github/workflows/trunk.yml

@@ -34,15 +34,6 @@ jobs:
       id-token: write
       contents: read
 
-  get-label-type:
-    name: get-label-type
-    uses: ./.github/workflows/_runner-determinator.yml
-    with:
-      triggering_actor: ${{ github.triggering_actor }}
-      issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
-      curr_branch: ${{ github.head_ref || github.ref_name }}
-      curr_ref_type: ${{ github.ref_type }}
-
   linux-focal-cuda12_4-py3_10-gcc9-sm86-build:
     name: linux-focal-cuda12.4-py3.10-gcc9-sm86
     uses: ./.github/workflows/_linux-build-label.yml
@@ -179,17 +170,15 @@ jobs:
   win-vs2019-cpu-py3-build:
     name: win-vs2019-cpu-py3
     uses: ./.github/workflows/_win-build.yml
-    needs: get-label-type
     with:
       build-environment: win-vs2019-cpu-py3
       cuda-version: cpu
       sync-tag: win-cpu-build
-      runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral"
       test-matrix: |
         { include: [
-          { config: "default", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral" },
-          { config: "default", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral" },
-          { config: "default", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral" },
+          { config: "default", shard: 1, num_shards: 3, runner: "windows.4xlarge.nonephemeral" },
+          { config: "default", shard: 2, num_shards: 3, runner: "windows.4xlarge.nonephemeral" },
+          { config: "default", shard: 3, num_shards: 3, runner: "windows.4xlarge.nonephemeral" },
         ]}
 
   win-vs2019-cpu-py3-test:
@@ -203,21 +192,28 @@ jobs:
       cuda-version: cpu
       test-matrix: ${{ needs.win-vs2019-cpu-py3-build.outputs.test-matrix }}
 
-  win-vs2019-cuda12_1-py3-build:
-    name: win-vs2019-cuda12.1-py3
+  win-vs2019-cuda11_8-py3-build:
+    name: win-vs2019-cuda11.8-py3
     uses: ./.github/workflows/_win-build.yml
-    needs: get-label-type
     with:
-      build-environment: win-vs2019-cuda12.1-py3
-      cuda-version: "12.1"
-      runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral"
+      build-environment: win-vs2019-cuda11.8-py3
+      cuda-version: "11.8"
+      sync-tag: win-cuda-build
+      test-matrix: |
+        { include: [
+          { config: "default", shard: 1, num_shards: 6, runner: "windows.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 2, num_shards: 6, runner: "windows.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 3, num_shards: 6, runner: "windows.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 4, num_shards: 6, runner: "windows.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 5, num_shards: 6, runner: "windows.g5.4xlarge.nvidia.gpu" },
+          { config: "default", shard: 6, num_shards: 6, runner: "windows.g5.4xlarge.nvidia.gpu" },
+          { config: "force_on_cpu", shard: 1, num_shards: 1, runner: "windows.4xlarge.nonephemeral" },
+        ]}
 
   linux-focal-rocm6_1-py3_8-build:
     name: linux-focal-rocm6.1-py3.8
     uses: ./.github/workflows/_linux-build-label.yml
-    needs: get-label-type
     with:
-      runner: "${{ needs.get-label-type.outputs.label-type }}linux.2xlarge"
       build-environment: linux-focal-rocm6.1-py3.8
       docker-image-name: pytorch-linux-focal-rocm-n-py3
       sync-tag: rocm-build
@@ -242,59 +238,3 @@ jobs:
       docker-image: ${{ needs.linux-focal-rocm6_1-py3_8-build.outputs.docker-image }}
       test-matrix: ${{ needs.linux-focal-rocm6_1-py3_8-build.outputs.test-matrix }}
       tests-to-include: "test_nn test_torch test_cuda test_ops test_unary_ufuncs test_binary_ufuncs test_autograd inductor/test_torchinductor distributed/test_c10d_common distributed/test_c10d_nccl"
-
-  linux-focal-cuda12_4-py3_10-gcc9-experimental-split-build:
-    name: linux-focal-cuda12.4-py3.10-gcc9-experimental-split-build
-    uses: ./.github/workflows/_linux-build-label.yml
-    with:
-      use_split_build: true
-      build-environment: linux-focal-cuda12.4-py3.10-gcc9
-      docker-image-name: pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9
-      test-matrix: |
-        { include: [
-          { config: "nogpu_AVX512", shard: 1, num_shards: 1, runner: "linux.2xlarge" },
-          { config: "nogpu_NO_AVX2", shard: 1, num_shards: 1, runner: "linux.2xlarge" },
-          { config: "jit_legacy", shard: 1, num_shards: 1, runner: "linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 1, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 2, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 3, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 4, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
-          { config: "default", shard: 5, num_shards: 5, runner: "linux.4xlarge.nvidia.gpu" },
-        ]}
-
-  linux-focal-cuda12_4-py3_10-gcc9-experimental-split-build-test:
-    name: linux-focal-cuda12.4-py3.10-gcc9-experimental-split-build-test
-    uses: ./.github/workflows/_linux-test.yml
-    needs:
-      - linux-focal-cuda12_4-py3_10-gcc9-experimental-split-build
-      - target-determination
-    with:
-      build-environment: linux-focal-cuda12.4-py3.10-gcc9-experimental-split-build
-      docker-image: ${{ needs.linux-focal-cuda12_4-py3_10-gcc9-experimental-split-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-focal-cuda12_4-py3_10-gcc9-experimental-split-build.outputs.test-matrix }}
-
-  linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build:
-    name: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build
-    uses: ./.github/workflows/_linux-build-label.yml
-    with:
-      use_split_build: true
-      build-environment: linux-focal-cuda11.8-py3.10-gcc9
-      docker-image-name: pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9
-      test-matrix: |
-        { include: [
-          { config: "distributed", shard: 1, num_shards: 3, runner: "linux.8xlarge.nvidia.gpu" },
-          { config: "distributed", shard: 2, num_shards: 3, runner: "linux.8xlarge.nvidia.gpu" },
-          { config: "distributed", shard: 3, num_shards: 3, runner: "linux.8xlarge.nvidia.gpu" },
-        ]}
-
-  linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build-test:
-    name: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build-test
-    uses: ./.github/workflows/_linux-test.yml
-    needs:
-      - linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build
-      - target-determination
-    with:
-      timeout-minutes: 360
-      build-environment: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build
-      docker-image: ${{ needs.linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build.outputs.test-matrix }}

.github/workflows/trymerge.yml

@@ -9,8 +9,6 @@ jobs:
     name: try_merge_pr_${{ github.event.client_payload.pr_num }}
     runs-on: linux.20_04.4x
     environment: mergebot
-    permissions:
-      id-token: write
     env:
         GH_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
     steps:
@@ -45,7 +43,6 @@ jobs:
           IGNORE_CURRENT: ${{ github.event.client_payload.ignore_current }}
           ROCKSET_API_KEY: ${{ secrets.ROCKSET_API_KEY }}
           DRCI_BOT_KEY: ${{ secrets.DRCI_BOT_KEY }}
-          GITHUB_RUN_ID: ${{ github.run_id }}
         run: |
           set -x
           if [ -n "${REBASE}" ]; then
@@ -87,22 +84,6 @@ jobs:
           set -x
           python3 .github/scripts/comment_on_pr.py "${PR_NUM}" "merge"
 
-      - name: configure aws credentials
-        uses: aws-actions/configure-aws-credentials@v3
-        continue-on-error: true
-        with:
-          role-to-assume: arn:aws:iam::308535385114:role/upload_to_ossci_raw_job_status
-          aws-region: us-east-1
-
-      - name: Upload merge record to s3
-        if: always()
-        continue-on-error: true
-        uses: seemethere/upload-artifact-s3@v5
-        with:
-          s3-bucket: ossci-raw-job-status
-          s3-prefix: merges/${{ github.repository }}/${{ github.event.client_payload.pr_num }}/${{ github.event.client_payload.comment_id }}/${{ github.run_id }}
-          path: merge_record.json
-
 # We want newer merge commands to supercede old ones
 concurrency:
   group: try-merge-${{ github.event.client_payload.pr_num }}

.github/workflows/upload-test-stats.yml

@@ -25,11 +25,12 @@ jobs:
 
   upload-test-stats:
     needs: get_workflow_conclusion
-    if: github.repository_owner == 'pytorch'
+    if:
+      github.repository_owner == 'pytorch' &&
+      (github.event.workflow_run.conclusion == 'success' || github.event.workflow_run.conclusion == 'failure' ||
+      needs.get_workflow_conclusion.outputs.conclusion == 'success' || needs.get_workflow_conclusion.outputs.conclusion == 'failure')
     runs-on: ubuntu-22.04
     environment: upload-stats
-    permissions:
-      id-token: write
     name: Upload test stats for ${{ github.event.workflow_run.id }}, attempt ${{ github.event.workflow_run.run_attempt }}
     steps:
       - name: Print workflow information
@@ -40,13 +41,6 @@ jobs:
       - name: Checkout PyTorch
         uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
 
-      - name: Configure aws credentials
-        uses: aws-actions/configure-aws-credentials@v3
-        continue-on-error: true
-        with:
-          role-to-assume: arn:aws:iam::308535385114:role/gha_workflow_upload-torch-test-stats
-          aws-region: us-east-1
-
       - uses: actions/setup-python@v4
         with:
           python-version: '3.11'
@@ -58,6 +52,8 @@ jobs:
       - name: Upload test artifacts
         id: upload-s3
         env:
+          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           WORKFLOW_ARTIFACTS_URL: ${{ github.event.workflow_run.artifacts_url }}
           WORKFLOW_RUN_ID: ${{ github.event.workflow_run.id }}
@@ -73,6 +69,8 @@ jobs:
       - name: Upload test stats
         env:
           ROCKSET_API_KEY: ${{ secrets.ROCKSET_API_KEY }}
+          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           WORKFLOW_RUN_ID: ${{ github.event.workflow_run.id }}
           WORKFLOW_RUN_ATTEMPT: ${{ github.event.workflow_run.run_attempt }}
@@ -86,6 +84,8 @@ jobs:
 
       - name: Analyze disabled tests rerun
         env:
+          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           WORKFLOW_ARTIFACTS_URL: ${{ github.event.workflow_run.artifacts_url }}
           WORKFLOW_RUN_ID: ${{ github.event.workflow_run.id }}
@@ -99,12 +99,14 @@ jobs:
         if: steps.upload-s3.outcome && steps.upload-s3.outcome == 'success' && github.event.workflow_run.name == 'inductor-micro-benchmark'
         env:
           ROCKSET_API_KEY: ${{ secrets.ROCKSET_API_KEY }}
+          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
           WORKFLOW_RUN_ID: ${{ github.event.workflow_run.id }}
           WORKFLOW_RUN_ATTEMPT: ${{ github.event.workflow_run.run_attempt }}
           REPO_FULLNAME: ${{ github.event.workflow_run.repository.full_name }}
           HEAD_BRANCH: ${{ github.event.workflow_run.head_branch }}
         run: |
-          python3 -m tools.stats.upload_dynamo_perf_stats --workflow-run-id "${WORKFLOW_RUN_ID}" --workflow-run-attempt "${WORKFLOW_RUN_ATTEMPT}" --repo "${REPO_FULLNAME}" --head-branch "${HEAD_BRANCH}" --rockset-collection oss_ci_benchmark --rockset-workspace benchmarks --dynamodb-table torchci-oss-ci-benchmark --match-filename "^gpt_fast_benchmark"
+          python3 -m tools.stats.upload_dynamo_perf_stats --workflow-run-id "${WORKFLOW_RUN_ID}" --workflow-run-attempt "${WORKFLOW_RUN_ATTEMPT}" --repo "${REPO_FULLNAME}" --head-branch "${HEAD_BRANCH}" --rockset-collection oss_ci_benchmark --rockset-workspace benchmarks --match-filename "^gpt_fast_benchmark"
 
   check-api-rate:
     if: ${{ always() && github.repository_owner == 'pytorch' }}

.github/workflows/upload-torch-dynamo-perf-stats.yml

@@ -26,8 +26,6 @@ jobs:
         github.event.workflow_run.conclusion == 'failure' || needs.get-conclusion.outputs.conclusion == 'failure'
     runs-on: ubuntu-22.04
     environment: upload-stats
-    permissions:
-      id-token: write
     name: Upload dynamo performance stats for ${{ github.event.workflow_run.id }}, attempt ${{ github.event.workflow_run.run_attempt }}
     steps:
       - name: Checkout PyTorch
@@ -36,13 +34,6 @@ jobs:
           submodules: false
           fetch-depth: 1
 
-      - name: Configure aws credentials
-        uses: aws-actions/configure-aws-credentials@v3
-        continue-on-error: true
-        with:
-          role-to-assume: arn:aws:iam::308535385114:role/gha_workflow_upload-torch-test-stats
-          aws-region: us-east-1
-
       - uses: actions/setup-python@v4
         with:
           python-version: '3.11'
@@ -54,6 +45,8 @@ jobs:
       - name: Upload torch dynamo performance stats to S3
         id: upload-s3
         env:
+          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           WORKFLOW_ARTIFACTS_URL: ${{ github.event.workflow_run.artifacts_url }}
           WORKFLOW_RUN_ID: ${{ github.event.workflow_run.id }}
@@ -68,9 +61,11 @@ jobs:
         if: steps.upload-s3.outcome && steps.upload-s3.outcome == 'success'
         env:
           ROCKSET_API_KEY: ${{ secrets.ROCKSET_API_KEY }}
+          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
           WORKFLOW_RUN_ID: ${{ github.event.workflow_run.id }}
           WORKFLOW_RUN_ATTEMPT: ${{ github.event.workflow_run.run_attempt }}
           REPO_FULLNAME: ${{ github.event.workflow_run.repository.full_name }}
           HEAD_BRANCH: ${{ github.event.workflow_run.head_branch }}
         run: |
-          python3 -m tools.stats.upload_dynamo_perf_stats --workflow-run-id "${WORKFLOW_RUN_ID}" --workflow-run-attempt "${WORKFLOW_RUN_ATTEMPT}" --repo "${REPO_FULLNAME}" --head-branch "${HEAD_BRANCH}" --rockset-collection torch_dynamo_perf_stats --rockset-workspace inductor --dynamodb-table torchci-dynamo-perf-stats --match-filename "^inductor_"
+          python3 -m tools.stats.upload_dynamo_perf_stats --workflow-run-id "${WORKFLOW_RUN_ID}" --workflow-run-attempt "${WORKFLOW_RUN_ATTEMPT}" --repo "${REPO_FULLNAME}" --head-branch "${HEAD_BRANCH}" --rockset-collection torch_dynamo_perf_stats --rockset-workspace inductor --match-filename "^inductor_"

.gitignore

@@ -129,7 +129,6 @@ env
 scripts/release_notes/*.json
 sccache-stats*.json
 lint.json
-merge_record.json
 
 # These files get copied over on invoking setup.py
 torchgen/packaged/*

.lintrunner.toml

@@ -68,7 +68,6 @@ include_patterns = [
     'aten/src/ATen/native/cudnn/*.cpp',
     'c10/**/*.h',
     'c10/**/*.cpp',
-    'distributed/c10d/*DMAConnectivity.*',
     'distributed/c10d/*SymmetricMemory.*',
     'torch/csrc/**/*.h',
     'torch/csrc/**/*.hpp',
@@ -204,8 +203,6 @@ include_patterns = [
     'torch/csrc/*.cpp',
     'torch/csrc/**/*.h',
     'torch/csrc/**/*.cpp',
-    'torch/csrc/jit/serialization/*.h',
-    'torch/csrc/jit/serialization/*.cpp',
 ]
 exclude_patterns = [
     # The negative filters below are to exclude files that include onnx_pb.h or
@@ -227,6 +224,7 @@ exclude_patterns = [
     'c10/util/win32-headers.h',
     'c10/util/*inl.h',
     'c10/test/**/*.h',
+    'aten/src/ATen/core/TensorImpl_test.cpp',
     'third_party/**/*',
     'torch/csrc/api/**',
     'torch/csrc/autograd/generated/**',
@@ -234,7 +232,8 @@ exclude_patterns = [
     'torch/csrc/dynamo/eval_frame.h',
     'torch/csrc/inductor/**/*',
     'torch/csrc/jit/**/*',
-    'torch/csrc/jit/serialization/mobile_bytecode_generated.h',
+    'torch/csrc/jit/serialization/import_legacy.cpp',
+    'torch/csrc/jit/serialization/export.cpp',
     'torch/csrc/lazy/**/*',
 ]
 init_command = [
@@ -1104,6 +1103,7 @@ exclude_patterns = [
     'test/test_prims.py',
     'test/test_proxy_tensor.py',
     'test/test_pruning_op.py',
+    'test/test_public_bindings.py',
     'test/test_quantization.py',
     'test/test_reductions.py',
     'test/test_scatter_gather_ops.py',
@@ -1129,6 +1129,8 @@ exclude_patterns = [
     'test/test_type_promotion.py',
     'test/test_unary_ufuncs.py',
     'test/test_vulkan.py',
+    'test/test_xnnpack_integration.py',
+    'test/torch_np/numpy_test/**/*.py',
     'torch/_awaits/__init__.py',
     'torch/_custom_op/__init__.py',
     'torch/_custom_op/autograd.py',
@@ -1189,6 +1191,9 @@ exclude_patterns = [
     'torch/_export/serde/upgrade.py',
     'torch/_export/trace.py',
     'torch/_export/verifier.py',
+    'torch/_higher_order_ops/__init__.py',
+    'torch/_higher_order_ops/out_dtype.py',
+    'torch/_higher_order_ops/wrap.py',
     'torch/_vendor/**',
     'torch/ao/__init__.py',
     'torch/ao/nn/__init__.py',
@@ -1531,7 +1536,12 @@ exclude_patterns = [
     'torch/signal/__init__.py',
     'torch/signal/windows/__init__.py',
     'torch/signal/windows/windows.py',
+    'torch/sparse/__init__.py',
+    'torch/sparse/_semi_structured_conversions.py',
+    'torch/sparse/_triton_ops.py',
+    'torch/sparse/semi_structured.py',
     'torch/special/__init__.py',
+    'torch/storage.py',
     'torch/testing/_internal/__init__.py',
     'torch/testing/_internal/autocast_test_lists.py',
     'torch/testing/_internal/autograd_function_db.py',
@@ -1604,6 +1614,7 @@ exclude_patterns = [
     'torch/testing/_internal/test_module/__init__.py',
     'torch/testing/_internal/test_module/future_div.py',
     'torch/testing/_internal/test_module/no_future_div.py',
+    'torch/utils/__init__.py',
     'torch/utils/_contextlib.py',
     'torch/utils/_cpp_extension_versioner.py',
     'torch/utils/_crash_handler.py',
@@ -1654,6 +1665,53 @@ exclude_patterns = [
     'torch/utils/collect_env.py',
     'torch/utils/cpp_backtrace.py',
     'torch/utils/cpp_extension.py',
+    'torch/utils/data/__init__.py',
+    'torch/utils/data/_utils/__init__.py',
+    'torch/utils/data/_utils/collate.py',
+    'torch/utils/data/_utils/fetch.py',
+    'torch/utils/data/_utils/pin_memory.py',
+    'torch/utils/data/_utils/serialization.py',
+    'torch/utils/data/_utils/signal_handling.py',
+    'torch/utils/data/_utils/worker.py',
+    'torch/utils/data/backward_compatibility.py',
+    'torch/utils/data/dataloader.py',
+    'torch/utils/data/datapipes/__init__.py',
+    'torch/utils/data/datapipes/_decorator.py',
+    'torch/utils/data/datapipes/_hook_iterator.py',
+    'torch/utils/data/datapipes/_typing.py',
+    'torch/utils/data/datapipes/dataframe/__init__.py',
+    'torch/utils/data/datapipes/dataframe/dataframe_wrapper.py',
+    'torch/utils/data/datapipes/dataframe/dataframes.py',
+    'torch/utils/data/datapipes/dataframe/datapipes.py',
+    'torch/utils/data/datapipes/dataframe/structures.py',
+    'torch/utils/data/datapipes/datapipe.py',
+    'torch/utils/data/datapipes/gen_pyi.py',
+    'torch/utils/data/datapipes/iter/__init__.py',
+    'torch/utils/data/datapipes/iter/callable.py',
+    'torch/utils/data/datapipes/iter/combinatorics.py',
+    'torch/utils/data/datapipes/iter/combining.py',
+    'torch/utils/data/datapipes/iter/filelister.py',
+    'torch/utils/data/datapipes/iter/fileopener.py',
+    'torch/utils/data/datapipes/iter/grouping.py',
+    'torch/utils/data/datapipes/iter/routeddecoder.py',
+    'torch/utils/data/datapipes/iter/selecting.py',
+    'torch/utils/data/datapipes/iter/sharding.py',
+    'torch/utils/data/datapipes/iter/streamreader.py',
+    'torch/utils/data/datapipes/iter/utils.py',
+    'torch/utils/data/datapipes/map/__init__.py',
+    'torch/utils/data/datapipes/map/callable.py',
+    'torch/utils/data/datapipes/map/combinatorics.py',
+    'torch/utils/data/datapipes/map/combining.py',
+    'torch/utils/data/datapipes/map/grouping.py',
+    'torch/utils/data/datapipes/map/utils.py',
+    'torch/utils/data/datapipes/utils/__init__.py',
+    'torch/utils/data/datapipes/utils/common.py',
+    'torch/utils/data/datapipes/utils/decoder.py',
+    'torch/utils/data/datapipes/utils/snapshot.py',
+    'torch/utils/data/distributed.py',
+    'torch/utils/data/graph.py',
+    'torch/utils/data/graph_settings.py',
+    'torch/utils/data/sampler.py',
     'torch/utils/dlpack.py',
     'torch/utils/file_baton.py',
     'torch/utils/flop_counter.py',
@@ -1693,9 +1751,8 @@ init_command = [
     '--dry-run={{DRYRUN}}',
     '--no-black-binary',
     'black==23.12.1',
-    'ufmt==2.7.0',
-    'usort==1.0.8.post1',
-    'isort==5.13.2',
+    'ufmt==2.1.0',
+    'usort==1.0.6',
 ]
 is_formatter = true
 
@@ -1779,7 +1836,7 @@ init_command = [
     'python3',
     'tools/linter/adapters/pip_init.py',
     '--dry-run={{DRYRUN}}',
-    'ruff==0.5.0',
+    'ruff==0.4.8',
 ]
 is_formatter = true
 

BUILD.bazel

@@ -762,7 +762,6 @@ cc_library(
         ":torch_headers",
         "@kineto",
         "@cpp-httplib",
-        "@nlohmann",
     ] + if_cuda([
         "@cuda//:nvToolsExt",
         "@cutlass",

CMakeLists.txt

@@ -208,6 +208,7 @@ endif()
 include(CMakeDependentOption)
 option(ATEN_NO_TEST "Do not build ATen test binaries" OFF)
 option(BUILD_BINARY "Build C++ binaries" OFF)
+option(BUILD_DOCS "Build Caffe2 documentation" OFF)
 option(BUILD_CUSTOM_PROTOBUF
        "Build and use Caffe2's own protobuf under third_party" ON)
 option(BUILD_PYTHON "Build Python binaries" ON)
@@ -749,6 +750,7 @@ if(NOT TORCH_BUILD_VERSION)
       CACHE STRING "Torch build version" FORCE)
 endif()
 caffe2_parse_version_str(TORCH ${TORCH_BUILD_VERSION})
+caffe2_parse_version_str(CAFFE2 ${TORCH_BUILD_VERSION})
 set(TORCH_SOVERSION "${TORCH_VERSION_MAJOR}.${TORCH_VERSION_MINOR}")
 
 # ---[ CMake scripts + modules
@@ -1221,6 +1223,45 @@ endif()
 add_subdirectory(c10)
 add_subdirectory(caffe2)
 
+# --[ Documentation
+if(BUILD_DOCS)
+  # check if Doxygen is installed
+  find_package(Doxygen)
+  if(DOXYGEN_FOUND)
+    message("Generating documentation")
+
+    set(DOXYGEN_C_IN ${CMAKE_CURRENT_SOURCE_DIR}/docs/caffe2/.Doxyfile-c)
+    set(DOXYGEN_C_OUT ${CMAKE_CURRENT_SOURCE_DIR}/docs/caffe2/Doxyfile-c)
+    set(DOXYGEN_P_IN ${CMAKE_CURRENT_SOURCE_DIR}/docs/caffe2/.Doxyfile-python)
+    set(DOXYGEN_P_OUT ${CMAKE_CURRENT_SOURCE_DIR}/docs/caffe2/Doxyfile-python)
+
+    if(EXISTS ${CMAKE_CURRENT_BINARY_DIR}/docs)
+      file(REMOVE_RECURSE ${CMAKE_CURRENT_BINARY_DIR}/docs)
+    endif()
+
+    file(MAKE_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/docs)
+    configure_file(${DOXYGEN_C_IN} ${DOXYGEN_C_OUT} @ONLY)
+    configure_file(${DOXYGEN_P_IN} ${DOXYGEN_P_OUT} @ONLY)
+
+    add_custom_target(
+      doc_doxygen_c ALL
+      COMMAND ${DOXYGEN_EXECUTABLE} ${DOXYGEN_C_OUT}
+      WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
+      COMMENT "Generating C++ API documentation with Doxygen"
+      VERBATIM)
+
+    add_custom_target(
+      doc_doxygen_python ALL
+      COMMAND ${DOXYGEN_EXECUTABLE} ${DOXYGEN_P_OUT}
+      WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
+      COMMENT "Generating Python API documentation with Doxygen"
+      VERBATIM)
+  else()
+    message(
+      FATAL_ERROR "Doxygen needs to be installed to generate the documentation")
+  endif()
+endif()
+
 # ---[ CMake related files Uninistall option.
 if(NOT TARGET caffe2_uninstall)
   configure_file(

CODEOWNERS

@@ -108,10 +108,10 @@ aten/src/ATen/detail/MTIAHooksInterface.h @egienvalue
 torch/csrc/mtia/ @egienvalue
 
 # Profiler
-torch/csrc/autograd/profiler* @aaronenyeshi @sraikund16
-torch/autograd/profiler* @aaronenyeshi @sraikund16
-torch/csrc/profiler/ @aaronenyeshi @sraikund16
-torch/profiler/ @aaronenyeshi @sraikund16
+torch/csrc/autograd/profiler* @aaronenyeshi
+torch/autograd/profiler* @aaronenyeshi
+torch/csrc/profiler/ @aaronenyeshi
+torch/profiler/ @aaronenyeshi
 
 # AOTDispatch tests
 test/functorch/test_aotdispatch.py @ezyang @Chillee

RELEASE.md

@@ -51,7 +51,6 @@ Following is the Release Compatibility Matrix for PyTorch releases:
 
 | PyTorch version | Python | Stable CUDA | Experimental CUDA | Stable ROCm |
 | --- | --- | --- | --- | --- |
-| 2.4 | >=3.8, <=3.12 | CUDA 11.8, CUDA 12.1, CUDNN 9.1.0.70  | CUDA 12.4, CUDNN 9.1.0.70 | ROCm 6.1 |
 | 2.3 | >=3.8, <=3.11, (3.12 experimental) | CUDA 11.8, CUDNN 8.7.0.84 | CUDA 12.1, CUDNN 8.9.2.26 | ROCm 6.0 |
 | 2.2 | >=3.8, <=3.11, (3.12 experimental) | CUDA 11.8, CUDNN 8.7.0.84 | CUDA 12.1, CUDNN 8.9.2.26 | ROCm 5.7 |
 | 2.1 | >=3.8, <=3.11 | CUDA 11.8, CUDNN 8.7.0.84 | CUDA 12.1, CUDNN 8.9.2.26 | ROCm 5.6 |
@@ -61,19 +60,15 @@ Following is the Release Compatibility Matrix for PyTorch releases:
 
 ## Release Cadence
 
-Following is the release cadence for year 2023/2024. All dates below are tentative, for latest updates on the release scheduled please follow [dev discuss](https://dev-discuss.pytorch.org/c/release-announcements/27). Please note: Patch Releases are optional.
+Following is the release cadence for year 2023/2024. All dates below are tentative, for latest updates on the release scheduled please follow [dev discuss](https://dev-discuss.pytorch.org/c/release-announcements/27).
 
 | Minor Version | Release branch cut | Release date | First patch release date | Second patch release date|
 | --- | --- | --- | --- | --- |
 | 2.1 | Aug 2023 | Oct 2023 | Nov 2023 | Dec 2023 |
 | 2.2 | Dec 2023 | Jan 2024 | Feb 2024 | Mar 2024 |
 | 2.3 | Mar 2024 | Apr 2024 | Jun 2024 | Not planned |
-| 2.4 | Jun 2024 | Jul 2024 | (Sept 2024) | Not planned |
-| 2.5 | Aug 2024 | Oct 2024 | (Nov 2024) | (Dec 2024) |
-| 2.6 | Dec 2024 | Jan 2025 | (Feb 2025) | (Mar 2025) |
-| 2.7 | Mar 2025 | Apr 2025 | (May 2025) | (Jun 2025) |
-| 2.8 | Jun 2025 | Jul 2025 | (Aug 2025) | (Sept 2025) |
-| 2.9 | Aug 2025 | Oct 2025 | (Nov 2025) | (Dec 2025) |
+| 2.4 | Jun 2024 | Jul 2024 | Aug 2024 | Sep 2024 |
+| 2.5 | Aug 2024 | Oct 2024 | Nov 2024 | Dec 2024 |
 
 ## General Overview
 

WORKSPACE

@@ -174,12 +174,6 @@ new_local_repository(
     path = "third_party/cpp-httplib",
 )
 
-new_local_repository(
-    name = "nlohmann",
-    build_file = "//third_party:nlohmann.BUILD",
-    path = "third_party/nlohmann",
-)
-
 new_local_repository(
     name = "tensorpipe",
     build_file = "//third_party:tensorpipe.BUILD",

android/pytorch_android/src/main/cpp/pytorch_jni_jit.cpp

@@ -119,7 +119,7 @@ class PytorchJni : public facebook::jni::HybridClass<PytorchJni> {
     }
     deviceType_ = deviceJniCodeToDeviceType(device);
     module_ = torch::jit::load(
-        std::move(modelPath->toStdString()), std::nullopt, extra_files);
+        std::move(modelPath->toStdString()), c10::nullopt, extra_files);
     if (has_extra) {
       static auto putMethod =
           facebook::jni::JMap<facebook::jni::JString, facebook::jni::JString>::

android/pytorch_android/src/main/cpp/pytorch_jni_lite.cpp

@@ -84,9 +84,9 @@ class PytorchJni : public facebook::jni::HybridClass<PytorchJni> {
     }
     deviceType_ = deviceJniCodeToDeviceType(device);
     module_ = torch::jit::_load_for_mobile(
-        std::move(modelPath->toStdString()), std::nullopt, extra_files);
+        std::move(modelPath->toStdString()), c10::nullopt, extra_files);
     torch::jit::_load_extra_only_for_mobile(
-        std::move(modelPath->toStdString()), std::nullopt, extra_files);
+        std::move(modelPath->toStdString()), c10::nullopt, extra_files);
     if (has_extra) {
       static auto putMethod =
           facebook::jni::JMap<facebook::jni::JString, facebook::jni::JString>::

aten/src/ATen/CMakeLists.txt

@@ -53,6 +53,11 @@ if(NOT BUILD_LITE_INTERPRETER)
   file(GLOB_RECURSE ATen_CORE_TEST_SRCS "core/*_test.cpp")
 endif()
 EXCLUDE(ATen_CORE_SRCS "${ATen_CORE_SRCS}" ${ATen_CORE_TEST_SRCS})
+# Exclude TensorImpl_test.cpp if compiling without Caffe2
+if(NOT BUILD_LITE_INTERPRETER)
+  file(GLOB_RECURSE ATen_CORE_EXCLUDED_TEST_SRCS "core/TensorImpl_test.cpp")
+  EXCLUDE(ATen_CORE_TEST_SRCS "${ATen_CORE_TEST_SRCS}" ${ATen_CORE_EXCLUDED_TEST_SRCS})
+endif()
 
 file(GLOB base_h "*.h" "detail/*.h" "cpu/*.h" "cpu/vec/vec512/*.h" "cpu/vec/vec256/*.h" "cpu/vec/vec256/vsx/*.h" "cpu/vec/vec256/zarch/*.h" "cpu/vec/*.h" "quantized/*.h" "functorch/*.h")
 file(GLOB base_cpp "*.cpp" "detail/*.cpp" "cpu/*.cpp" "functorch/*.cpp")

aten/src/ATen/CPUGeneratorImpl.cpp

@@ -222,7 +222,7 @@ c10::intrusive_ptr<c10::TensorImpl> CPUGeneratorImpl::get_state() const {
   static const size_t size = sizeof(CPUGeneratorImplState);
   static_assert(std::is_standard_layout_v<CPUGeneratorImplState>, "CPUGeneratorImplState is not a PODType");
 
-  auto state_tensor = at::detail::empty_cpu({(int64_t)size}, ScalarType::Byte, std::nullopt, std::nullopt, std::nullopt, std::nullopt);
+  auto state_tensor = at::detail::empty_cpu({(int64_t)size}, ScalarType::Byte, c10::nullopt, c10::nullopt, c10::nullopt, c10::nullopt);
   auto rng_state = state_tensor.data_ptr();
 
   // accumulate generator data to be copied into byte tensor

aten/src/ATen/CPUGeneratorImpl.h

@@ -3,7 +3,7 @@
 #include <ATen/core/Generator.h>
 #include <ATen/core/MT19937RNGEngine.h>
 #include <c10/core/GeneratorImpl.h>
-#include <optional>
+#include <c10/util/Optional.h>
 
 namespace at {
 

aten/src/ATen/Context.cpp

@@ -56,14 +56,6 @@ void Context::setDeterministicCuDNN(bool b) {
   deterministic_cudnn = b;
 }
 
-bool Context::deterministicMkldnn() const {
-  return deterministic_mkldnn;
-}
-
-void Context::setDeterministicMkldnn(bool b) {
-  deterministic_mkldnn = b;
-}
-
 bool Context::deterministicAlgorithms() const {
   return _deterministic_algorithms;
 }
@@ -153,13 +145,6 @@ void Context::setSDPUseCuDNN(bool e) {
   enabled_cudnnSDP = e;
 }
 
-void Context::setSDPUseOverrideable(bool e) {
-  enabled_overrideable = e;
-}
-
-bool Context::userEnabledOverrideableSDP() const {
-  return enabled_overrideable;
-}
 
 // NOLINTNEXTLINE(cppcoreguidelines-avoid-c-arrays,modernize-avoid-c-arrays)
 static const char cublas_config_var_name[] = "CUBLAS_WORKSPACE_CONFIG";
@@ -278,24 +263,7 @@ void Context::setLinalgPreferredBackend(at::LinalgBackend b) {
   }
 }
 
-at::BlasBackend Context::blasPreferredBackend() {
-#ifdef USE_ROCM
-  if (blas_preferred_backend == at::BlasBackend::Cublaslt) {
-    static const bool hipblaslt_unsupported = []() {
-      static const std::vector<std::string> archs = {"gfx90a", "gfx940", "gfx941", "gfx942"};
-      for (auto index: c10::irange(getNumGPUs())) {
-        if (!detail::getCUDAHooks().isGPUArch(index, archs)) {
-          TORCH_WARN_ONCE(
-            "Attempting to use hipBLASLt on an unsupported architecture! "
-            "Overriding blas backend to hipblas");
-          return true;
-        }
-      }
-      return false;
-    }();
-    if (hipblaslt_unsupported) blas_preferred_backend = at::BlasBackend::Cublas;
-  }
-#endif
+at::BlasBackend Context::blasPreferredBackend() const {
   return blas_preferred_backend;
 }
 

aten/src/ATen/Context.h

@@ -59,14 +59,12 @@ class TORCH_API Context {
     }
   }
   const AcceleratorHooksInterface& getAcceleratorHooksInterface(
-      std::optional<c10::DeviceType> opt_device_type = std::nullopt) {
+      std::optional<c10::DeviceType> opt_device_type = c10::nullopt) {
     c10::DeviceType device_type = opt_device_type.has_value()
         ? opt_device_type.value()
         : at::getAccelerator(true).value();
     if (device_type == at::kCUDA) {
       return at::detail::getCUDAHooks();
-    } else if (device_type == at::kXPU) {
-      return at::detail::getXPUHooks();
     } else if (device_type == at::kMPS) {
       return at::detail::getMPSHooks();
     } else if (device_type == at::kPrivateUse1) {
@@ -190,8 +188,6 @@ class TORCH_API Context {
   void setBenchmarkLimitCuDNN(int);
   bool deterministicCuDNN() const;
   void setDeterministicCuDNN(bool);
-  bool deterministicMkldnn() const;
-  void setDeterministicMkldnn(bool);
   bool userEnabledNNPACK() const;
   void setUserEnabledNNPACK(bool e);
 
@@ -218,13 +214,10 @@ class TORCH_API Context {
   void setSDPUseCuDNN(bool);
   bool userEnabledCuDNNSDP() const;
 
-  void setSDPUseOverrideable(bool);
-  bool userEnabledOverrideableSDP() const;
-
   at::LinalgBackend linalgPreferredBackend() const;
   void setLinalgPreferredBackend(at::LinalgBackend);
 
-  at::BlasBackend blasPreferredBackend();
+  at::BlasBackend blasPreferredBackend() const;
   void setBlasPreferredBackend(at::BlasBackend);
 
   // Note [Enabling Deterministic Operations]
@@ -365,15 +358,13 @@ class TORCH_API Context {
   c10::once_flag thp_init;
   bool enabled_cudnn = true;
   bool deterministic_cudnn = false;
-  bool deterministic_mkldnn = false;
   bool _deterministic_algorithms = false;
   bool _deterministic_algorithms_warn_only = false;
   bool _deterministic_fill_uninitialized_memory = true;
   bool enabled_flashSDP = true;
   bool enabled_mem_efficientSDP = true;
   bool enabled_mathSDP = true;
-  bool enabled_cudnnSDP = true;
-  bool enabled_overrideable = true;
+  bool enabled_cudnnSDP = false;
 #ifdef USE_ROCM
   bool benchmark_cudnn = true;
 #else
@@ -407,7 +398,7 @@ class TORCH_API Context {
   bool release_original_weights = false;
 #endif
   bool display_vmap_fallback_warnings_ = false;
-  std::optional<at::QEngine> quantized_engine = std::nullopt;
+  std::optional<at::QEngine> quantized_engine = c10::nullopt;
   bool enable_sparse_tensor_invariant_checks = false;
   bool allow_fp16_reduction_cpu = false;
 

aten/src/ATen/DLConvertor.cpp

@@ -115,9 +115,6 @@ static DLDevice getDLDevice(const Tensor& tensor, c10::DeviceIndex device_id) {
       ctx.device_id =
           at::detail::getXPUHooks().getGlobalIdxFromDevice(tensor.device());
       break;
-    case DeviceType::MAIA:
-      ctx.device_type = DLDeviceType::kDLMAIA;
-      break;
     default:
       TORCH_CHECK(false, "Cannot pack tensors on " + tensor.device().str());
   }
@@ -144,8 +141,6 @@ static Device getATenDevice(const DLDevice& ctx, void* data) {
 #endif
     case DLDeviceType::kDLOneAPI:
       return at::detail::getXPUHooks().getDeviceFromPtr(data);
-    case DLDeviceType::kDLMAIA:
-      return at::Device(DeviceType::MAIA, ctx.device_id);
     default:
       TORCH_CHECK(
           false, "Unsupported device_type: ", std::to_string(ctx.device_type));

aten/src/ATen/DeviceAccelerator.cpp

@@ -1,37 +1,39 @@
-#include <ATen/Context.h>
 #include <ATen/DeviceAccelerator.h>
+#include <ATen/Context.h>
+
 namespace at {
 
 C10_API std::optional<DeviceType> getAccelerator(bool checked) {
-#define DETECT_AND_ASSIGN_ACCELERATOR(device_name) \
-  if (at::has##device_name()) {                    \
-    device_type = k##device_name;                  \
-    TORCH_CHECK(                                   \
-        !is_accelerator_detected,                  \
-        "Cannot have ",                            \
-        device_type.value(),                       \
-        " with other accelerators.");              \
-    is_accelerator_detected = true;                \
-  }
+#define CHECK_NO_CUDA \
+  TORCH_CHECK(!at::hasCUDA(), "Cannot have both CUDA and PrivateUse1");
 
-  if (is_privateuse1_backend_registered()) {
-    // We explicitly allow PrivateUse1 and another device at the same time as we
-    // use this for testing. Whenever a PrivateUse1 device is registered, use it
-    // first.
-    return kPrivateUse1;
-  }
-  std::optional<DeviceType> device_type = std::nullopt;
-  bool is_accelerator_detected = false;
-  DETECT_AND_ASSIGN_ACCELERATOR(CUDA)
-  DETECT_AND_ASSIGN_ACCELERATOR(MTIA)
-  DETECT_AND_ASSIGN_ACCELERATOR(XPU)
-  if (checked) {
-    TORCH_CHECK(
-        device_type, "Cannot access accelerator device when none is available.")
-  }
-  return device_type;
+#define CHECK_NO_PU1 \
+  TORCH_CHECK(!is_privateuse1_backend_registered(), "Cannot have both CUDA and PrivateUse1");
 
-#undef DETECT_AND_ASSIGN_ACCELERATOR
+#define CHECK_NO_MTIA \
+  TORCH_CHECK(!at::hasMTIA(), "Cannot have MTIA with other devices");
+
+    if (is_privateuse1_backend_registered()) {
+        // We explicitly allow PrivateUse1 and another device at the same time
+        // as we use this for testing.
+        // Whenever a PrivateUse1 device is registered, use it first.
+        return kPrivateUse1;
+    } else if (at::hasCUDA()) {
+        CHECK_NO_PU1
+        CHECK_NO_MTIA
+        return kCUDA;
+    } else if (at::hasMTIA()) {
+        CHECK_NO_CUDA
+        CHECK_NO_PU1
+        return kMTIA;
+    } else {
+        TORCH_CHECK(!checked, "Cannot access accelerator device when none is available.")
+        return std::nullopt;
+    }
+
+#undef CHECK_NO_CUDA
+#undef CHECK_NO_PU1
 }
 
+
 } // namespace at

aten/src/ATen/DeviceAccelerator.h

@@ -13,9 +13,9 @@
 // - It provides a set of common APIs as defined by AcceleratorHooksInterface
 //
 // As of today, accelerator devices are (in no particular order):
-// CUDA, MTIA, XPU, PrivateUse1
+// CUDA, MTIA, PrivateUse1
 // We want to add once all the proper APIs are supported and tested:
-// HIP, MPS
+// HIP, MPS, XPU
 
 namespace at {
 

aten/src/ATen/DeviceGuard.h

@@ -17,14 +17,14 @@ namespace at {
 /// Return the Device of a Tensor, if the Tensor is defined.
 inline std::optional<Device> device_of(const Tensor& t) {
   if (t.defined()) {
-    return std::make_optional(t.device());
+    return c10::make_optional(t.device());
   } else {
-    return std::nullopt;
+    return c10::nullopt;
   }
 }
 
 inline std::optional<Device> device_of(const std::optional<Tensor>& t) {
-  return t.has_value() ? device_of(t.value()) : std::nullopt;
+  return t.has_value() ? device_of(t.value()) : c10::nullopt;
 }
 
 /// Return the Device of a TensorList, if the list is non-empty and
@@ -34,7 +34,7 @@ inline std::optional<Device> device_of(ITensorListRef t) {
   if (!t.empty()) {
     return device_of(t.front());
   } else {
-    return std::nullopt;
+    return c10::nullopt;
   }
 }
 

aten/src/ATen/EmptyTensor.cpp

@@ -29,7 +29,6 @@ c10::Allocator* GetCPUAllocatorMaybePinned(bool pin_memory) {
   return c10::GetCPUAllocator();
 }
 
-#ifndef C10_MOBILE
 constexpr uint64_t storage_max() {
   // int64_t and size_t are used somewhat inconsistently throughout ATen.
   // To be safe, storage size calculations must fit in both types.
@@ -39,7 +38,6 @@ constexpr uint64_t storage_max() {
       std::numeric_limits<size_t>::max());
   return std::min(int64_max, size_max);
 }
-#endif
 
 inline void raise_warning_for_complex_half(ScalarType dtype) {
   if (dtype == kComplexHalf) {

aten/src/ATen/EmptyTensor.h

@@ -76,7 +76,7 @@ TORCH_API TensorBase empty_cpu(
     IntArrayRef size,
     ScalarType dtype,
     bool pin_memory = false,
-    std::optional<c10::MemoryFormat> memory_format_opt = std::nullopt);
+    std::optional<c10::MemoryFormat> memory_format_opt = c10::nullopt);
 
 TORCH_API TensorBase empty_cpu(
     IntArrayRef size,
@@ -110,7 +110,7 @@ TORCH_API TensorBase empty_strided_cpu(
 TORCH_API TensorBase empty_meta(
     IntArrayRef size,
     ScalarType dtype,
-    std::optional<c10::MemoryFormat> memory_format_opt = std::nullopt);
+    std::optional<c10::MemoryFormat> memory_format_opt = c10::nullopt);
 
 TORCH_API TensorBase empty_meta(
     IntArrayRef size,

aten/src/ATen/FunctionalInverses.cpp

@@ -303,7 +303,7 @@ Tensor FunctionalInverses::_nested_view_from_buffer_inverse(const Tensor& base,
     return Tensor();
 }
 
-Tensor FunctionalInverses::_nested_view_from_jagged_inverse(const Tensor& base, const Tensor& mutated_view, InverseReturnMode inverse_return_mode, const Tensor& offsets, const Tensor& dummy, const std::optional<Tensor>& lengths, int64_t ragged_idx, const std::optional<Tensor>& min_seqlen, const std::optional<Tensor>& max_seqlen) {
+Tensor FunctionalInverses::_nested_view_from_jagged_inverse(const Tensor& base, const Tensor& mutated_view, InverseReturnMode inverse_return_mode, const Tensor& offsets, const Tensor& dummy, const std::optional<Tensor>& lengths, int64_t ragged_idx) {
   auto values = at::_nested_get_values(mutated_view);
   if (inverse_return_mode != InverseReturnMode::NeverView) {
     return values;
@@ -317,12 +317,7 @@ Tensor FunctionalInverses::_nested_get_values_inverse(const Tensor& base, const
   auto lengths = at::_nested_get_lengths(base);
   auto ragged_idx = at::_nested_get_ragged_idx(base);
   auto dummy = at::_nested_get_jagged_dummy(base);
-  auto min_seqlen = at::_nested_get_min_seqlen(base);
-  auto max_seqlen = at::_nested_get_max_seqlen(base);
-  auto nt = at::_nested_view_from_jagged(
-      mutated_view, offsets, dummy, lengths, ragged_idx,
-      (min_seqlen.defined() ? std::optional<Tensor>(min_seqlen) : std::nullopt),
-      (max_seqlen.defined() ? std::optional<Tensor>(max_seqlen) : std::nullopt));
+  auto nt = at::_nested_view_from_jagged(mutated_view, offsets, dummy, lengths, ragged_idx);
 
   if (inverse_return_mode != InverseReturnMode::NeverView) {
     return nt;

aten/src/ATen/FunctionalTensorWrapper.cpp

@@ -514,9 +514,6 @@ c10::SymInt FunctionalTensorWrapper::sym_size_custom(int64_t d) const {
 c10::SymInt FunctionalTensorWrapper::sym_storage_offset_custom() const {
   return value_.unsafeGetTensorImpl()->sym_storage_offset();
 }
-c10::Layout FunctionalTensorWrapper::layout_impl() const {
-  return value_.unsafeGetTensorImpl()->layout();
-}
 
 namespace functionalization {
 namespace impl {
@@ -531,9 +528,9 @@ Tensor to_functional_tensor(const Tensor& tensor) {
 }
 std::optional<Tensor> to_functional_tensor(const std::optional<Tensor>& tensor) {
   if (tensor.has_value()) {
-    return std::make_optional<Tensor>(to_functional_tensor(*tensor));
+    return c10::make_optional<Tensor>(to_functional_tensor(*tensor));
   }
-  return std::nullopt;
+  return c10::nullopt;
 }
 c10::List<::std::optional<Tensor>> to_functional_tensor(const c10::List<::std::optional<Tensor>>& t_list) {
   c10::List<::std::optional<Tensor>> outputs;
@@ -569,9 +566,9 @@ Tensor from_functional_tensor(const Tensor& tensor, bool assert_functional) {
 }
 std::optional<Tensor> from_functional_tensor(const std::optional<Tensor>& t, bool assert_functional) {
   if (t.has_value()) {
-    return std::make_optional<Tensor>(from_functional_tensor(*t, assert_functional));
+    return c10::make_optional<Tensor>(from_functional_tensor(*t, assert_functional));
   }
-  return std::nullopt;
+  return c10::nullopt;
 }
 std::vector<Tensor> from_functional_tensor(ITensorListRef t_list) {
   std::vector<Tensor> outputs;

aten/src/ATen/FunctionalTensorWrapper.h

@@ -222,7 +222,6 @@ struct TORCH_API FunctionalTensorWrapper : public c10::TensorImpl {
   c10::SymIntArrayRef sym_strides_custom() const override;
   c10::SymInt sym_storage_offset_custom() const override;
   c10::Device device_custom() const override;
-  c10::Layout layout_impl() const override;
 
  private:
   const char* tensorimpl_type_name() const override;

aten/src/ATen/FunctionalizeFallbackKernel.cpp

@@ -217,7 +217,7 @@ static at::Tensor lift_fresh_functionalize_copy(const at::Tensor & self) {
     // we will end up hitting PreDispatch stack first. So, we should
     // directly redispatch to the functionalize key manually.
     static auto op = c10::Dispatcher::singleton().findSchemaOrThrow("aten::clone", "").typed<at::Tensor(const at::Tensor &, std::optional<at::MemoryFormat>)>();
-    return op.redispatch(c10::DispatchKeySet({c10::DispatchKey::Functionalize}), self, std::nullopt);
+    return op.redispatch(c10::DispatchKeySet({c10::DispatchKey::Functionalize}), self, c10::nullopt);
   }
 
   at::AutoDispatchSkipFunctionalize guard;

aten/src/ATen/InferSize.h

@@ -4,7 +4,7 @@
 #include <c10/core/ScalarType.h>
 #include <c10/core/SymIntArrayRef.h>
 #include <c10/util/DimVector.h>
-#include <optional>
+#include <c10/util/Optional.h>
 #include <sstream>
 #include <vector>
 

aten/src/ATen/LegacyBatchedFallback.cpp

@@ -139,7 +139,7 @@ static void batchedTensorInplaceForLoopFallback(const c10::OperatorHandle& op, t
     if (self_vmap_levels != (self_vmap_levels | other_vmap_levels)) {
       // Find one vmap level to complain about
       auto additional_bdims = (self_vmap_levels | other_vmap_levels) ^ self_vmap_levels;
-      [[maybe_unused]] auto offending_level = llvm::findLastSet(additional_bdims.to_ulong());
+      auto offending_level = llvm::findLastSet(additional_bdims.to_ulong());
       // The following prints out "vmap: aten::add_(tensor, ...) is not possible",
       // but it would be better to print out "tensor.add_(...) is not possible".
       // Afaict there's no official way to get the add_ and there is no way to

aten/src/ATen/LegacyBatchingRegistrations.cpp

@@ -62,7 +62,7 @@ static bool is_allowed_dim_on_scalar_tensor(int64_t dim) {
   return dim == 0 || dim == -1;
 }
 
-Tensor sum_batching_rule(const Tensor& self, OptionalIntArrayRef opt_dims, bool keepdim, std::optional<ScalarType> dtype) {
+Tensor sum_batching_rule(const Tensor& self, OptionalIntArrayRef opt_dims, bool keepdim, optional<ScalarType> dtype) {
   if (opt_dims.has_value()) {
     auto dims = opt_dims.value();
     // PyTorch has a special case where sum(scalar_tensor, dim=0) does not fail
@@ -198,7 +198,7 @@ std::vector<Tensor> chunk_batching_rule(const Tensor& self, int64_t chunks, int6
   return result;
 }
 
-Tensor clamp_batching_rule(const Tensor& self, const std::optional<Scalar>& min, const std::optional<Scalar>& max) {
+Tensor clamp_batching_rule(const Tensor& self, const optional<Scalar>& min, const optional<Scalar>& max) {
   auto self_physical = MultiBatchVmapTransform::logicalToPhysical(self);
   auto result = at::clamp(self_physical.tensor(), min, max);
   return self_physical.getPhysicalToLogicalMap().apply(result);
@@ -508,11 +508,11 @@ static void checkBatchDimsAtFrontInLayout(IntArrayRef physical_strides, int64_t
 // given (sizes, strides, storage_offset) returns the maximum location that
 // can be indexed (or nullopt if such a location doesn't exist, e.g., tensors
 // with zero-size dims).
-static std::optional<int64_t> maximum_indexable_location(
+static optional<int64_t> maximum_indexable_location(
     IntArrayRef sizes, IntArrayRef strides, int64_t storage_offset) {
   auto result = native::storage_size_for(sizes, strides);
   if (result == 0) {
-    return std::nullopt;
+    return nullopt;
   }
   return result + storage_offset;
 }
@@ -526,7 +526,7 @@ static void checkBasicAsStridedValidForSlice(
     int64_t num_batch_dims,
     IntArrayRef sizes,
     IntArrayRef strides,
-    std::optional<int64_t> maybe_storage_offset) {
+    optional<int64_t> maybe_storage_offset) {
   auto slice_sizes = physical_tensor.sizes().slice(num_batch_dims);
   auto slice_strides = physical_tensor.strides().slice(num_batch_dims);
   auto base_offset = physical_tensor.storage_offset();
@@ -614,7 +614,7 @@ Tensor as_strided_batching_rule(
     const Tensor& tensor,
     IntArrayRef sizes,
     IntArrayRef strides,
-    std::optional<int64_t> storage_offset) {
+    optional<int64_t> storage_offset) {
   auto physical_view = at::MultiBatchVmapTransform::logicalToPhysical(tensor);
   auto num_batch_dims = physical_view.numBatchDims();
   auto physical_sizes = physical_view.getPhysicalShape(sizes);
@@ -763,7 +763,7 @@ Tensor pow_scalar_Tensor_batching_rule(const Scalar& other, const Tensor& self)
   return makeBatched(output_physical, BatchDims(old_bdims.begin(), old_bdims.end()));
 }
 
-Tensor clone_batching_rule(const Tensor& self, std::optional<MemoryFormat> memory_format) {
+Tensor clone_batching_rule(const Tensor& self, optional<MemoryFormat> memory_format) {
   // Memory format support is a little tricky because vmap is allowed to move
   // around batch dimensions and some memory formats are rank-dependent.
   // Another weird case is:
@@ -958,12 +958,12 @@ Tensor stack_batching_rule(TensorList tensors, int64_t dim) {
 // unwrap_and_call<..., at::to> because at::to takes TensorOptions& (!!)
 Tensor to_dtype_layout_batching_rule(
     const Tensor& self,
-    std::optional<ScalarType> dtype,
-    std::optional<Layout> layout,
-    std::optional<Device> device,
-    std::optional<bool> pin_memory,
+    optional<ScalarType> dtype,
+    optional<Layout> layout,
+    optional<Device> device,
+    optional<bool> pin_memory,
     bool non_blocking, bool copy,
-    std::optional<MemoryFormat> memory_format) {
+    optional<MemoryFormat> memory_format) {
   auto options = TensorOptions()
     .dtype(dtype)
     .layout(layout)
@@ -978,10 +978,10 @@ Tensor to_dtype_layout_batching_rule(
 Tensor new_zeros_batching_rule(
     const Tensor& self,
     IntArrayRef size,
-    std::optional<ScalarType> dtype,
-    std::optional<Layout> layout,
-    std::optional<Device> device,
-    std::optional<bool> pin_memory) {
+    optional<ScalarType> dtype,
+    optional<Layout> layout,
+    optional<Device> device,
+    optional<bool> pin_memory) {
   auto physical_view = MultiBatchVmapTransform::logicalToPhysical(self);
   auto physical_size = physical_view.getPhysicalShape(size);
   auto options = TensorOptions()
@@ -1010,10 +1010,10 @@ Tensor new_empty_strided_batching_rule(
     const Tensor& self,
     IntArrayRef size,
     IntArrayRef stride,
-    std::optional<ScalarType> dtype,
-    std::optional<Layout> layout,
-    std::optional<Device> device,
-    std::optional<bool> pin_memory) {
+    optional<ScalarType> dtype,
+    optional<Layout> layout,
+    optional<Device> device,
+    optional<bool> pin_memory) {
   auto physical_view = MultiBatchVmapTransform::logicalToPhysical(self);
   auto physical_size = physical_view.getPhysicalShape(size);
 
@@ -1181,9 +1181,9 @@ TORCH_LIBRARY_IMPL(aten, Batched, m) {
     m.impl(name, unwrap_and_call_method< \
         to_type, &Tensor::to, __VA_ARGS__>);\
   }
-  TO_BATCHING_RULE("to.device", Device, ScalarType, bool, bool, std::optional<MemoryFormat>)
-  TO_BATCHING_RULE("to.dtype", ScalarType, bool, bool, std::optional<MemoryFormat>)
-  TO_BATCHING_RULE("to.other", const Tensor&, bool, bool, std::optional<MemoryFormat>)
+  TO_BATCHING_RULE("to.device", Device, ScalarType, bool, bool, optional<MemoryFormat>)
+  TO_BATCHING_RULE("to.dtype", ScalarType, bool, bool, optional<MemoryFormat>)
+  TO_BATCHING_RULE("to.other", const Tensor&, bool, bool, optional<MemoryFormat>)
   m.impl("to.dtype_layout", to_dtype_layout_batching_rule);
 #undef TO_BATCHING_RULE
   m.impl("clone", clone_batching_rule);

aten/src/ATen/MapAllocator.cpp

@@ -275,9 +275,7 @@ MapAllocator::MapAllocator(WithFd, c10::string_view filename, int fd, int flags,
 
     struct stat file_stat{};
     if (fstat(fd, &file_stat) == -1) {
-#ifndef STRIP_ERROR_MESSAGES
       int last_err = errno;
-#endif
       if (!(flags_ & ALLOCATOR_MAPPED_FROMFD)) {
         ::close(fd);
       }
@@ -291,9 +289,7 @@ MapAllocator::MapAllocator(WithFd, c10::string_view filename, int fd, int flags,
             TORCH_CHECK(false, "unable to resize file <", filename_, "> to the right size: ", strerror(errno), " (", errno, ")");
           }
           if (fstat(fd, &file_stat) == -1 || file_stat.st_size < static_cast<int64_t>(size)) {
-#ifndef STRIP_ERROR_MESSAGES
             int last_err = errno;
-#endif
             ::close(fd);
             TORCH_CHECK(false, "unable to stretch file <", filename_, "> to the right size: ", strerror(last_err), " (", last_err, ")");
           }
@@ -302,9 +298,7 @@ MapAllocator::MapAllocator(WithFd, c10::string_view filename, int fd, int flags,
  */
 #ifndef __APPLE__
           if ((write(fd, "", 1)) != 1) /* note that the string "" contains the '\0' byte ... */ {
-#ifndef STRIP_ERROR_MESSAGES
             int last_err = errno;
-#endif
             ::close(fd);
             TORCH_CHECK(false, "unable to write to file <", filename_, ">: ", strerror(last_err), " (", last_err, ")");
           }

aten/src/ATen/NamedTensorUtils.cpp

@@ -8,14 +8,12 @@
 
 namespace at {
 
-#ifndef STRIP_ERROR_MESSAGES
 // Returns "Tensor['N', 'C', 'H', 'W']" for a tensor with names ('N', 'C', 'H', 'W').
 static std::string toDimnameRepr(const Tensor& tensor) {
   std::ostringstream os;
   os << "Tensor" << tensor.names();
   return os.str();
 }
-#endif
 
 int64_t dimname_to_position(const Tensor& tensor, Dimname dim) {
   TORCH_CHECK(dim.type() != NameType::WILDCARD,

aten/src/ATen/NestedTensorImpl.cpp

@@ -173,7 +173,7 @@ NestedTensorImpl::NestedTensorImpl(
       nested_sizes_(std::move(nested_sizes)),
       nested_strides_(std::move(nested_strides)),
       storage_offsets_(std::move(storage_offsets)),
-      opt_sizes_(std::nullopt) {
+      opt_sizes_(c10::nullopt) {
   C10_LOG_API_USAGE_ONCE("torch.NestedTensor");
   TORCH_WARN_ONCE(
       "The PyTorch API of nested tensors is in prototype stage and will change "
@@ -230,7 +230,7 @@ NestedTensorImpl::NestedTensorImpl(
       nested_sizes_(std::move(nested_sizes)),
       nested_strides_(std::move(nested_strides)),
       storage_offsets_(std::move(storage_offsets)),
-      opt_sizes_(std::nullopt) {
+      opt_sizes_(c10::nullopt) {
   validate_nested_tensor_metadata(nested_sizes_, nested_strides_, storage_offsets_);
   refresh_dim();
   set_custom_sizes_strides(c10::TensorImpl::SizesStridesPolicy::CustomSizes);
@@ -239,11 +239,11 @@ NestedTensorImpl::NestedTensorImpl(
 std::optional<int64_t> NestedTensorImpl::opt_size(int64_t d) const {
   if (C10_UNLIKELY(!opt_sizes_.has_value())) {
     // Cache the metadata to avoid recomputing it each time.
-    opt_sizes_ = std::make_optional(construct_opt_sizes(nested_sizes_));
+    opt_sizes_ = c10::make_optional(construct_opt_sizes(nested_sizes_));
   }
   d = at::maybe_wrap_dim(d, dim(), false);
   if ((*opt_sizes_)[d] == -1) {
-    return std::nullopt;
+    return c10::nullopt;
   }
   return (*opt_sizes_)[d];
 }

aten/src/ATen/ParallelCommon.cpp

@@ -29,7 +29,6 @@ const char* get_env_var(
   return value ? value : def_value;
 }
 
-#ifndef C10_MOBILE
 size_t get_env_num_threads(const char* var_name, size_t def_value = 0) {
   try {
     if (auto* value = std::getenv(var_name)) {
@@ -44,7 +43,6 @@ size_t get_env_num_threads(const char* var_name, size_t def_value = 0) {
   }
   return def_value;
 }
-#endif
 
 } // namespace
 

aten/src/ATen/SavedTensorHooks.cpp

@@ -32,7 +32,7 @@ void SavedTensorDefaultHooks::disable(const std::string& message) {
 }
 
 void SavedTensorDefaultHooks::enable() {
-  tls.disabled_error_message = std::nullopt;
+  tls.disabled_error_message = c10::nullopt;
 }
 
 /* static */ bool SavedTensorDefaultHooks::set_tracing(bool is_tracing) {

aten/src/ATen/SavedTensorHooks.h

@@ -1,8 +1,8 @@
 #pragma once
 
 #include <c10/macros/Export.h>
+#include <c10/util/Optional.h>
 #include <c10/util/python_stub.h>
-#include <optional>
 #include <stack>
 #include <string>
 

aten/src/ATen/ScalarOps.cpp

@@ -27,7 +27,7 @@ Tensor scalar_tensor_static(const Scalar& s, std::optional<ScalarType> dtype_opt
   at::tracer::impl::NoTracerDispatchMode tracer_guard;
   at::AutoDispatchBelowAutograd mode;
   Tensor result = at::detail::empty_cpu(
-      {}, dtype_opt, std::nullopt, device_opt, std::nullopt, std::nullopt);
+      {}, dtype_opt, c10::nullopt, device_opt, c10::nullopt, c10::nullopt);
   scalar_fill(result, s);
   return result;
 }

aten/src/ATen/SparseTensorImpl.h

@@ -140,7 +140,7 @@ struct TORCH_API SparseTensorImpl : public TensorImpl {
         "), but got ",
         size.size());
     if (nnz() > 0) {
-      [[maybe_unused]] auto constexpr alt_options_msg =
+      auto alt_options_msg =
           "You could try the following options:\n\
 1. If you need an empty sparse tensor of this size, call `x = torch.sparse_coo_tensor(size)`.\n\
 2. If you need to resize this tensor, you have the following options:\n\

aten/src/ATen/TensorIndexing.h

@@ -5,8 +5,8 @@
 #include <ATen/core/Tensor.h>
 #include <ATen/core/TensorBody.h>
 #include <c10/core/SymInt.h>
+#include <c10/util/Optional.h>
 #include <c10/util/irange.h>
-#include <optional>
 
 #ifndef AT_PER_OPERATOR_HEADERS
 #include <ATen/Functions.h>
@@ -29,7 +29,7 @@ constexpr int64_t INDEX_MAX = -(INDEX_MIN + 1);
 
 enum class TensorIndexType { None, Ellipsis, SymInt, Boolean, Slice, Tensor };
 
-constexpr std::nullopt_t None = std::nullopt;
+constexpr c10::nullopt_t None = c10::nullopt;
 
 struct TORCH_API EllipsisIndexType final {
   EllipsisIndexType() = default;
@@ -39,9 +39,9 @@ TORCH_API extern const EllipsisIndexType Ellipsis;
 struct TORCH_API Slice final {
  public:
   Slice(
-      std::optional<c10::SymInt> start_index = std::nullopt,
-      std::optional<c10::SymInt> stop_index = std::nullopt,
-      std::optional<c10::SymInt> step_index = std::nullopt) {
+      std::optional<c10::SymInt> start_index = c10::nullopt,
+      std::optional<c10::SymInt> stop_index = c10::nullopt,
+      std::optional<c10::SymInt> step_index = c10::nullopt) {
     if (!step_index.has_value()) {
       step_ = c10::SymInt(1);
     } else {
@@ -110,7 +110,7 @@ TORCH_API std::ostream& operator<<(std::ostream& stream, const Slice& slice);
 // `torch.tensor([1, 2])`) | `torch::tensor({1, 2})`
 struct TORCH_API TensorIndex final {
   // Case 1: `at::indexing::None`
-  TensorIndex(std::nullopt_t) : type_(TensorIndexType::None) {}
+  TensorIndex(c10::nullopt_t) : type_(TensorIndexType::None) {}
 
   // Case 2: "..." / `at::indexing::Ellipsis`
   TensorIndex(at::indexing::EllipsisIndexType)
@@ -530,7 +530,7 @@ inline Tensor applySlicing(
     auto& obj = indices[i];
     // See NOTE [nested tensor size for indexing]
     std::optional<SymIntArrayRef> result_sizes = result.is_nested()
-        ? std::optional<SymIntArrayRef>(std::nullopt)
+        ? std::optional<SymIntArrayRef>(c10::nullopt)
         : std::optional<SymIntArrayRef>(result.sym_sizes());
     result = handleDimInMultiDimIndexing(
         /*prev_dim_result=*/result,
@@ -606,7 +606,7 @@ inline Tensor get_item(
   // as null may need to be changed after we reach a better solution for nested
   // tensor size
   std::optional<SymIntArrayRef> self_sizes = self.is_nested()
-      ? std::optional<SymIntArrayRef>(std::nullopt)
+      ? std::optional<SymIntArrayRef>(c10::nullopt)
       : std::optional<SymIntArrayRef>(self.sym_sizes());
 
   // handle simple types: integers, slices, none, ellipsis, bool

aten/src/ATen/TensorIterator.cpp

@@ -171,7 +171,7 @@ TensorIteratorConfig& TensorIteratorConfig::declare_static_shape(IntArrayRef sha
   //   This will bypass all shape checking in the TensorIterator. Kernels which call this method
   //   are expected to check shapes before calling `add_owned_input` or `add_owned_output`.
   TORCH_CHECK(!resize_outputs_, "resize_outputs() must be called before declare_static_shape(...)")
-  static_shape_ = std::make_optional(DimVector(shape));
+  static_shape_ = c10::make_optional(DimVector(shape));
   return *this;
 }
 

aten/src/ATen/TensorIterator.h

@@ -147,7 +147,7 @@ struct TORCH_API OperandInfo {
   /// promotion target_dtype value can become different from tensor's dtype
   /// also, during type promotion target_dtype and device can be set for an
   /// undefined tensor so that tensor can be properly constructed later.
-  std::optional<Device> device = std::nullopt;
+  std::optional<Device> device = c10::nullopt;
   ScalarType target_dtype = ScalarType::Undefined;
   // Caches dtype of the tensor, because scalar_type is an expensive operation
   // If dtype of the tensor is changed (e.g. as a result of type promotion or in
@@ -971,9 +971,9 @@ class TORCH_API TensorIteratorConfig final {
   int num_outputs_ = 0;
   int num_inputs_ = 0;
 
-  std::optional<DimVector> static_shape_ = std::nullopt;
-  std::optional<ScalarType> static_dtype_ = std::nullopt;
-  std::optional<Device> static_device_ = std::nullopt;
+  std::optional<DimVector> static_shape_ = c10::nullopt;
+  std::optional<ScalarType> static_dtype_ = c10::nullopt;
+  std::optional<Device> static_device_ = c10::nullopt;
   bool check_mem_overlap_ = true;
   bool allow_cpu_scalars_ = false;
   bool is_reduction_ = false;