chore: rebase

Author: peri044

.github/scripts/filter-matrix.py

@@ -3,8 +3,9 @@
 import argparse
 import json
 import sys
+from typing import List
 
-disabled_python_versions = "3.13"
+disabled_python_versions: List[str] = []
 
 
 def main(args: list[str]) -> None:

.github/scripts/generate-tensorrt-test-matrix.py

@@ -28,6 +28,10 @@
 # please update the future tensorRT version you want to test here
 TENSORRT_VERSIONS_DICT = {
     "windows": {
+        "10.3.0": {
+            "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.3.0/zip/TensorRT-10.3.0.26.Windows.win10.cuda-12.5.zip",
+            "strip_prefix": "TensorRT-10.3.0.26",
+        },
         "10.7.0": {
             "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.7.0/zip/TensorRT-10.7.0.23.Windows.win10.cuda-12.6.zip",
             "strip_prefix": "TensorRT-10.7.0.23",
@@ -42,6 +46,10 @@
         },
     },
     "linux": {
+        "10.3.0": {
+            "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.3.0/tars/TensorRT-10.3.0.26.Linux.x86_64-gnu.cuda-12.5.tar.gz",
+            "strip_prefix": "TensorRT-10.3.0.26",
+        },
         "10.7.0": {
             "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.7.0/tars/TensorRT-10.7.0.23.Linux.x86_64-gnu.cuda-12.6.tar.gz",
             "strip_prefix": "TensorRT-10.7.0.23",

.github/scripts/generate_binary_build_matrix.py

@@ -18,15 +18,16 @@
 import sys
 from typing import Any, Callable, Dict, List, Optional, Tuple
 
+PYTHON_VERSIONS_FOR_PR_BUILD = ["3.11"]
 PYTHON_ARCHES_DICT = {
-    "nightly": ["3.9", "3.10", "3.11", "3.12"],
-    "test": ["3.9", "3.10", "3.11", "3.12"],
-    "release": ["3.9", "3.10", "3.11", "3.12"],
+    "nightly": ["3.9", "3.10", "3.11", "3.12", "3.13"],
+    "test": ["3.9", "3.10", "3.11", "3.12", "3.13"],
+    "release": ["3.9", "3.10", "3.11", "3.12", "3.13"],
 }
 CUDA_ARCHES_DICT = {
     "nightly": ["11.8", "12.6", "12.8"],
     "test": ["11.8", "12.6", "12.8"],
-    "release": ["11.8", "12.6", "12.8"],
+    "release": ["11.8", "12.4", "12.6"],
 }
 ROCM_ARCHES_DICT = {
     "nightly": ["6.1", "6.2"],
@@ -422,11 +423,6 @@ def generate_wheels_matrix(
         # Define default python version
         python_versions = list(PYTHON_ARCHES)
 
-        # If the list of python versions is set explicitly by the caller, stick with it instead
-        # of trying to add more versions behind the scene
-        if channel == NIGHTLY and (os in (LINUX, MACOS_ARM64, LINUX_AARCH64)):
-            python_versions += ["3.13"]
-
     if os == LINUX:
         # NOTE: We only build manywheel packages for linux
         package_type = "manywheel"
@@ -456,7 +452,7 @@ def generate_wheels_matrix(
             arches += [XPU]
 
     if limit_pr_builds:
-        python_versions = [python_versions[0]]
+        python_versions = PYTHON_VERSIONS_FOR_PR_BUILD
 
     global WHEEL_CONTAINER_IMAGES
 

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

@@ -23,7 +23,6 @@ jobs:
       test-infra-ref: main
       with-rocm: false
       with-cpu: false
-      python-versions: '["3.11", "3.12", "3.10", "3.9"]'
 
   filter-matrix:
     needs: [generate-matrix]

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

@@ -23,7 +23,6 @@ jobs:
       test-infra-ref: main
       with-rocm: false
       with-cpu: false
-      python-versions: '["3.11", "3.12", "3.10", "3.9"]'
 
   substitute-runner:
     needs: generate-matrix

.github/workflows/docgen.yml

@@ -35,7 +35,7 @@ jobs:
             - name: Install base deps
               run: |
                   python3 -m pip install pip --upgrade
-                  python3 -m pip install pyyaml numpy torch --pre --extra-index-url https://download.pytorch.org/whl/nightly/cu126
+                  python3 -m pip install pyyaml numpy torch --pre --extra-index-url https://download.pytorch.org/whl/nightly/cu128
                   ./packaging/pre_build_script.sh
             - name: Get HEAD SHA
               id: vars
@@ -44,7 +44,7 @@ jobs:
               env:
                   USE_PRE_CXX11_ABI: 0
               run: |
-                  python3 -m pip install --pre . --extra-index-url https://download.pytorch.org/whl/nightly/cu126
+                  python3 -m pip install --pre . --extra-index-url https://download.pytorch.org/whl/nightly/cu128
             - name: Generate New Docs
               run: |
                   cd docsrc

MODULE.bazel

@@ -55,21 +55,21 @@ http_archive(
     name = "libtorch",
     build_file = "@//third_party/libtorch:BUILD",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/nightly/cu128/libtorch-cxx11-abi-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/test/cu128/libtorch-cxx11-abi-shared-with-deps-latest.zip"],
 )
 
 http_archive(
     name = "libtorch_pre_cxx11_abi",
     build_file = "@//third_party/libtorch:BUILD",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/nightly/cu128/libtorch-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/test/cu126/libtorch-shared-with-deps-latest.zip"],
 )
 
 http_archive(
     name = "libtorch_win",
     build_file = "@//third_party/libtorch:BUILD",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/nightly/cu128/libtorch-win-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/test/cu128/libtorch-win-shared-with-deps-latest.zip"],
 )
 
 # Download these tarballs manually from the NVIDIA website

core/runtime/TRTEngine.cpp

@@ -453,6 +453,10 @@ std::vector<std::string> TRTEngine::serialize() {
   return serialized_info;
 }
 
+void TRTEngine::reset_captured_graph() {
+  cudagraph.reset();
+}
+
 } // namespace runtime
 } // namespace core
 } // namespace torch_tensorrt

core/runtime/TRTEngine.h

@@ -185,6 +185,7 @@ struct TRTEngine : torch::CustomClassHolder {
   // c10::List<at::Tensor> Run(c10::List<at::Tensor> inputs);
 
   void set_profiling_paths();
+  void reset_captured_graph();
 #ifndef NDEBUG
   bool profile_execution = true;
 #else

core/runtime/register_jit_hooks.cpp

@@ -88,6 +88,7 @@ static auto TORCHTRT_UNUSED TRTEngineTSRegistrtion =
         .def("dump_engine_layer_info", &TRTEngine::dump_engine_layer_info)
         .def("get_engine_layer_info", &TRTEngine::get_engine_layer_info)
         .def("infer_outputs", &TRTEngine::infer_outputs)
+        .def("reset_captured_graph", &TRTEngine::reset_captured_graph)
         .def_readwrite("use_pre_allocated_outputs", &TRTEngine::use_pre_allocated_outputs)
         .def_readwrite("use_output_allocator_outputs", &TRTEngine::use_output_allocator_outputs)
         .def_property(