diff --git a/.circleci/scripts/binary_populate_env.sh b/.circleci/scripts/binary_populate_env.sh
index 4cec3e2f6d72..aa82d36aa7ce 100755
--- a/.circleci/scripts/binary_populate_env.sh
+++ b/.circleci/scripts/binary_populate_env.sh
@@ -75,14 +75,7 @@ export PYTORCH_BUILD_NUMBER=1
 : <<'BLOCK_COMMENT'
 # Set triton version as part of PYTORCH_EXTRA_INSTALL_REQUIREMENTS
 TRITON_VERSION=$(cat $PYTORCH_ROOT/.ci/docker/triton_version.txt)
-
-# Here PYTORCH_EXTRA_INSTALL_REQUIREMENTS is already set for the all the wheel builds hence append TRITON_CONSTRAINT
-TRITON_CONSTRAINT="platform_system == 'Linux' and platform_machine == 'x86_64'"
-
-# CUDA 12.9/13.0 builds have triton for Linux and Linux aarch64 binaries.
-if [[ "$DESIRED_CUDA" == "cu129" ]] || [[ "$DESIRED_CUDA" == "cu130" ]]; then
-  TRITON_CONSTRAINT="platform_system == 'Linux'"
-fi
+TRITON_CONSTRAINT="platform_system == 'Linux'"
 
 if [[ "$PACKAGE_TYPE" =~ .*wheel.* &&  -n "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:-}" && ! "$PYTORCH_BUILD_VERSION" =~ .*xpu.* ]]; then
   TRITON_REQUIREMENT="triton==${TRITON_VERSION}; ${TRITON_CONSTRAINT}"
diff --git a/.github/templates/linux_binary_build_workflow.yml.j2 b/.github/templates/linux_binary_build_workflow.yml.j2
index f53472571993..bf7db5866e78 100644
--- a/.github/templates/linux_binary_build_workflow.yml.j2
+++ b/.github/templates/linux_binary_build_workflow.yml.j2
@@ -77,6 +77,9 @@ jobs:
       runs_on: linux.s390x
       ALPINE_IMAGE: "docker.io/s390x/alpine"
       timeout-minutes: 420
+      {%- elif config["gpu_arch_type"] == "rocm" %}
+      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       {%- elif "conda" in build_environment and config["gpu_arch_type"] == "cuda" %}
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.24xlarge.ephemeral
diff --git a/.github/workflows/generated-linux-binary-libtorch-nightly.yml b/.github/workflows/generated-linux-binary-libtorch-nightly.yml
index f5eca8751840..bc671ae80ae2 100644
--- a/.github/workflows/generated-linux-binary-libtorch-nightly.yml
+++ b/.github/workflows/generated-linux-binary-libtorch-nightly.yml
@@ -333,6 +333,7 @@ jobs:
       LIBTORCH_CONFIG: release
       LIBTORCH_VARIANT: shared-with-deps
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: libtorch-rocm6_3-shared-with-deps-release
       build_environment: linux-binary-libtorch
     secrets:
@@ -446,6 +447,7 @@ jobs:
       LIBTORCH_CONFIG: release
       LIBTORCH_VARIANT: shared-with-deps
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: libtorch-rocm6_4-shared-with-deps-release
       build_environment: linux-binary-libtorch
     secrets:
diff --git a/.github/workflows/generated-linux-binary-manywheel-nightly.yml b/.github/workflows/generated-linux-binary-manywheel-nightly.yml
index c996437a3b9f..5f9eaab976a6 100644
--- a/.github/workflows/generated-linux-binary-manywheel-nightly.yml
+++ b/.github/workflows/generated-linux-binary-manywheel-nightly.yml
@@ -323,6 +323,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.3
       DESIRED_PYTHON: "3.10"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_10-rocm6_3
       build_environment: linux-binary-manywheel
     secrets:
@@ -433,6 +434,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.4
       DESIRED_PYTHON: "3.10"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_10-rocm6_4
       build_environment: linux-binary-manywheel
     secrets:
@@ -912,6 +914,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.3
       DESIRED_PYTHON: "3.11"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_11-rocm6_3
       build_environment: linux-binary-manywheel
     secrets:
@@ -1022,6 +1025,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.4
       DESIRED_PYTHON: "3.11"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_11-rocm6_4
       build_environment: linux-binary-manywheel
     secrets:
@@ -1501,6 +1505,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.3
       DESIRED_PYTHON: "3.12"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_12-rocm6_3
       build_environment: linux-binary-manywheel
     secrets:
@@ -1611,6 +1616,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.4
       DESIRED_PYTHON: "3.12"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_12-rocm6_4
       build_environment: linux-binary-manywheel
     secrets:
@@ -2090,6 +2096,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.3
       DESIRED_PYTHON: "3.13"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_13-rocm6_3
       build_environment: linux-binary-manywheel
     secrets:
@@ -2200,6 +2207,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.4
       DESIRED_PYTHON: "3.13"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_13-rocm6_4
       build_environment: linux-binary-manywheel
     secrets:
@@ -2679,6 +2687,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.3
       DESIRED_PYTHON: "3.13t"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_13t-rocm6_3
       build_environment: linux-binary-manywheel
     secrets:
@@ -2789,6 +2798,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.4
       DESIRED_PYTHON: "3.13t"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_13t-rocm6_4
       build_environment: linux-binary-manywheel
     secrets:
@@ -3268,6 +3278,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.3
       DESIRED_PYTHON: "3.14"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_14-rocm6_3
       build_environment: linux-binary-manywheel
     secrets:
@@ -3378,6 +3389,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.4
       DESIRED_PYTHON: "3.14"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_14-rocm6_4
       build_environment: linux-binary-manywheel
     secrets:
@@ -3857,6 +3869,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.3
       DESIRED_PYTHON: "3.14t"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_14t-rocm6_3
       build_environment: linux-binary-manywheel
     secrets:
@@ -3967,6 +3980,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.4
       DESIRED_PYTHON: "3.14t"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_14t-rocm6_4
       build_environment: linux-binary-manywheel
     secrets:
diff --git a/.github/workflows/generated-linux-binary-manywheel-rocm-main.yml b/.github/workflows/generated-linux-binary-manywheel-rocm-main.yml
index 9593391217ac..9df4835757c4 100644
--- a/.github/workflows/generated-linux-binary-manywheel-rocm-main.yml
+++ b/.github/workflows/generated-linux-binary-manywheel-rocm-main.yml
@@ -60,6 +60,7 @@ jobs:
       DOCKER_IMAGE_TAG_PREFIX: rocm6.4
       DESIRED_PYTHON: "3.10"
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
+      timeout-minutes: 300
       build_name: manywheel-py3_10-rocm6_4
       build_environment: linux-binary-manywheel-rocm
     secrets:
diff --git a/.github/workflows/operator_benchmark.yml b/.github/workflows/operator_benchmark.yml
index dd262d31b8fc..dcdc2cd0ba24 100644
--- a/.github/workflows/operator_benchmark.yml
+++ b/.github/workflows/operator_benchmark.yml
@@ -14,6 +14,10 @@ on:
   schedule:
     # Run at 07:00 UTC every Sunday
     - cron: 0 7 * * 0
+  pull_request:
+    paths:
+      - benchmarks/operator_benchmark/**
+      - .github/workflows/operator_benchmark.yml
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name == 'workflow_dispatch' }}
diff --git a/Dockerfile b/Dockerfile
index f73dfcc1af3a..331cf00593cb 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -53,7 +53,7 @@ ARG CUDA_PATH=cu121
 ARG INSTALL_CHANNEL=whl/nightly
 # Automatically set by buildx
 # pinning version of conda here see: https://github.com/pytorch/pytorch/issues/164574
-RUN /opt/conda/bin/conda install -c "${INSTALL_CHANNEL}" -y python=${PYTHON_VERSION} conda=25.7.0
+RUN /opt/conda/bin/conda install -y python=${PYTHON_VERSION} conda=25.7.0
 
 ARG TARGETPLATFORM
 
diff --git a/aten/src/ATen/mps/EmptyTensor.cpp b/aten/src/ATen/mps/EmptyTensor.cpp
index d858df073397..6c58de099648 100644
--- a/aten/src/ATen/mps/EmptyTensor.cpp
+++ b/aten/src/ATen/mps/EmptyTensor.cpp
@@ -12,7 +12,7 @@
 
 #define MPS_ERROR_NOT_COMPILED "PyTorch code is not compiled with MPS enabled"
 #define MPS_ERROR_RUNTIME_TOO_LOW \
-  "The MPS backend is supported on MacOS 13.0+.", \
+  "The MPS backend is supported on MacOS 14.0+. ", \
   "Current OS version can be queried using `sw_vers`"
 #define MPS_ERROR_DOUBLE_NOT_SUPPORTED "Cannot convert a MPS Tensor to float64 dtype " \
   "as the MPS framework doesn't support float64. Please use float32 instead."
diff --git a/benchmarks/operator_benchmark/expected_ci_operator_benchmark_eager_float32_cpu.csv b/benchmarks/operator_benchmark/expected_ci_operator_benchmark_eager_float32_cpu.csv
index 873f14d20127..9a7b6797e982 100644
--- a/benchmarks/operator_benchmark/expected_ci_operator_benchmark_eager_float32_cpu.csv
+++ b/benchmarks/operator_benchmark/expected_ci_operator_benchmark_eager_float32_cpu.csv
@@ -1,5 +1,5 @@
 Benchmarking Framework,Benchmarking Module Name,Case Name,tag,run_backward,Execution Time
-PyTorch,add,add_M1_N1_K1_cpu,short,FALSE,3.9497
+PyTorch,add,add_M1_N1_K1_cpu,short,FALSE,2.459
 PyTorch,add,add_M64_N64_K64_cpu,short,FALSE,14.3181
 PyTorch,add,add_M64_N64_K128_cpu,short,FALSE,14.6826
 PyTorch,add,add_M1_N1_K1_cpu_bwdall_BACKWARD,short,TRUE,58.1449
@@ -376,10 +376,10 @@ PyTorch,relu6,"relu6_dims(3,4,5)_contigFalse_inplaceFalse_dtypetorch.qint32",sho
 PyTorch,relu6,"relu6_dims(2,3,4,5)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,9.6588
 PyTorch,relu6,"relu6_dims(2,3,4,5)_contigFalse_inplaceFalse_dtypetorch.qint8",short,FALSE,9.5969
 PyTorch,relu6,"relu6_dims(2,3,4,5)_contigFalse_inplaceFalse_dtypetorch.qint32",short,FALSE,9.547
-PyTorch,relu6,"relu6_dims(512,512)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,68.739
+PyTorch,relu6,"relu6_dims(512,512)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,50.21375
 PyTorch,relu6,"relu6_dims(512,512)_contigFalse_inplaceFalse_dtypetorch.qint8",short,FALSE,45.14133333
 PyTorch,relu6,"relu6_dims(512,512)_contigFalse_inplaceFalse_dtypetorch.qint32",short,FALSE,52.6664
-PyTorch,relu6,"relu6_dims(256,1024)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,69.1875
+PyTorch,relu6,"relu6_dims(256,1024)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,51.49525
 PyTorch,relu6,"relu6_dims(256,1024)_contigFalse_inplaceFalse_dtypetorch.qint8",short,FALSE,48.3458
 PyTorch,relu6,"relu6_dims(256,1024)_contigFalse_inplaceFalse_dtypetorch.qint32",short,FALSE,62.0719
 PyTorch,functional.hardtanh,"functional.hardtanh_dims(3,4,5)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,7.5728
@@ -388,10 +388,10 @@ PyTorch,functional.hardtanh,"functional.hardtanh_dims(3,4,5)_contigFalse_inplace
 PyTorch,functional.hardtanh,"functional.hardtanh_dims(2,3,4,5)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,8.1647
 PyTorch,functional.hardtanh,"functional.hardtanh_dims(2,3,4,5)_contigFalse_inplaceFalse_dtypetorch.qint8",short,FALSE,8.1768
 PyTorch,functional.hardtanh,"functional.hardtanh_dims(2,3,4,5)_contigFalse_inplaceFalse_dtypetorch.qint32",short,FALSE,8.0619
-PyTorch,functional.hardtanh,"functional.hardtanh_dims(512,512)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,67.118
+PyTorch,functional.hardtanh,"functional.hardtanh_dims(512,512)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,48.88475
 PyTorch,functional.hardtanh,"functional.hardtanh_dims(512,512)_contigFalse_inplaceFalse_dtypetorch.qint8",short,FALSE,43.702
 PyTorch,functional.hardtanh,"functional.hardtanh_dims(512,512)_contigFalse_inplaceFalse_dtypetorch.qint32",short,FALSE,50.3613
-PyTorch,functional.hardtanh,"functional.hardtanh_dims(256,1024)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,67.436
+PyTorch,functional.hardtanh,"functional.hardtanh_dims(256,1024)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,50.3995
 PyTorch,functional.hardtanh,"functional.hardtanh_dims(256,1024)_contigFalse_inplaceFalse_dtypetorch.qint8",short,FALSE,46.9813
 PyTorch,functional.hardtanh,"functional.hardtanh_dims(256,1024)_contigFalse_inplaceFalse_dtypetorch.qint32",short,FALSE,59.2295
 PyTorch,functional.hardsigmoid,"functional.hardsigmoid_dims(3,4,5)_contigFalse_inplaceFalse_dtypetorch.quint8",short,FALSE,6.5189
@@ -1316,4 +1316,4 @@ PyTorch,where,"where_cond_shape(8,16,1)_input_shape(1,)_other_shape(1,)_cpu_dtyp
 PyTorch,where,"where_cond_shape(8,16,1)_input_shape(16,1)_other_shape(8,16,1)_cpu_dtypetorch.float32",short,FALSE,5.763
 PyTorch,where,"where_cond_shape(8,16,1)_input_shape(8,1,1)_other_shape(1,)_cpu_dtypetorch.float32",short,FALSE,5.744666667
 PyTorch,clamp,clamp_M512_N512_cpu,short,FALSE,15.26233333
-PyTorch,gelu,gelu_M512_N512_cpu,short,FALSE,31.33166667
\ No newline at end of file
+PyTorch,gelu,gelu_M512_N512_cpu,short,FALSE,31.33166667
diff --git a/torch/__init__.py b/torch/__init__.py
index a5c072396e1d..0625ad60bfff 100644
--- a/torch/__init__.py
+++ b/torch/__init__.py
@@ -302,7 +302,7 @@ def _get_cuda_dep_paths(path: str, lib_folder: str, lib_name: str) -> list[str]:
     return nvidia_lib_paths + lib_paths
 
 
-def _preload_cuda_deps(lib_folder: str, lib_name: str) -> None:
+def _preload_cuda_deps(lib_folder: str, lib_name: str, required: bool = True) -> None:  # type: ignore[valid-type]
     """Preloads cuda deps if they could not be found otherwise."""
     # Should only be called on Linux if default path resolution have failed
     assert platform.system() == "Linux", "Should only be called on Linux"
@@ -313,9 +313,10 @@ def _preload_cuda_deps(lib_folder: str, lib_name: str) -> None:
         if candidate_lib_paths:
             lib_path = candidate_lib_paths[0]
             break
-    if not lib_path:
+    if not lib_path and required:
         raise ValueError(f"{lib_name} not found in the system path {sys.path}")
-    ctypes.CDLL(lib_path)
+    if lib_path:
+        ctypes.CDLL(lib_path)
 
 
 # See Note [Global dependencies]
@@ -354,8 +355,6 @@ def _load_global_deps() -> None:
     except OSError as err:
         # Can only happen for wheel with cuda libs as PYPI deps
         # As PyTorch is not purelib, but nvidia-*-cu12 is
-        from torch.version import cuda as cuda_version
-
         cuda_libs: dict[str, str] = {
             "cublas": "libcublas.so.*[0-9]",
             "cudnn": "libcudnn.so.*[0-9]",
@@ -369,7 +368,6 @@ def _load_global_deps() -> None:
             "cusparselt": "libcusparseLt.so.*[0-9]",
             "cusolver": "libcusolver.so.*[0-9]",
             "nccl": "libnccl.so.*[0-9]",
-            "nvtx": "libnvToolsExt.so.*[0-9]",
             "nvshmem": "libnvshmem_host.so.*[0-9]",
             "cufile": "libcufile.so.*[0-9]",
         }
@@ -381,6 +379,9 @@ def _load_global_deps() -> None:
             raise err
         for lib_folder, lib_name in cuda_libs.items():
             _preload_cuda_deps(lib_folder, lib_name)
+
+        # libnvToolsExt is Optional Dependency
+        _preload_cuda_deps("nvtx", "libnvToolsExt.so.*[0-9]", required=False)
         ctypes.CDLL(global_deps_lib_path, mode=ctypes.RTLD_GLOBAL)
 
 
diff --git a/torch/utils/cpp_extension.py b/torch/utils/cpp_extension.py
index 7202a9638756..902d2fe6ce0f 100644
--- a/torch/utils/cpp_extension.py
+++ b/torch/utils/cpp_extension.py
@@ -2418,10 +2418,6 @@ def _get_cuda_arch_flags(cflags: Optional[list[str]] = None) -> list[str]:
 
     # If not given or set as native, determine what's best for the GPU / CUDA version that can be found
     if not _arch_list or _arch_list == "native":
-        if not _arch_list:
-            logger.warning(
-                "TORCH_CUDA_ARCH_LIST is not set, all archs for visible cards are included for compilation. \n"
-                "If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST'] to specific architectures.")
         arch_list = []
         # the assumption is that the extension should run on any of the currently visible cards,
         # which could be of different types - therefore all archs for visible cards should be included
@@ -2440,6 +2436,15 @@ def _get_cuda_arch_flags(cflags: Optional[list[str]] = None) -> list[str]:
                 arch_list.append(arch)
         arch_list = sorted(arch_list)
         arch_list[-1] += '+PTX'
+
+        if not _arch_list:
+            # Only log on rank 0 in distributed settings to avoid spam
+            if not torch.distributed.is_available() or not torch.distributed.is_initialized() or torch.distributed.get_rank() == 0:
+                arch_list_str = ';'.join(arch_list)
+                logger.debug(
+                    "TORCH_CUDA_ARCH_LIST is not set, using TORCH_CUDA_ARCH_LIST='%s' "
+                    "for visible GPU architectures. Set os.environ['TORCH_CUDA_ARCH_LIST'] to override.",
+                    arch_list_str)
     else:
         # Deal with lists that are ' ' separated (only deal with ';' after)
         _arch_list = _arch_list.replace(' ', ';')