Upgrade to TensorRT 10.11 (#3557)

Author: lanluo-nvidia

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

@@ -44,6 +44,14 @@
             "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.9.0/zip/TensorRT-10.9.0.34.Windows.win10.cuda-12.8.zip",
             "strip_prefix": "TensorRT-10.9.0.34",
         },
+        "10.10.0": {
+            "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.10.0/zip/TensorRT-10.10.0.31.Windows.win10.cuda-12.9.zip",
+            "strip_prefix": "TensorRT-10.10.0.31",
+        },
+        "10.11.0": {
+            "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.11.0/zip/TensorRT-10.11.0.33.Windows.win10.cuda-12.9.zip",
+            "strip_prefix": "TensorRT-10.11.0.33",
+        },
     },
     "linux": {
         "10.3.0": {
@@ -62,13 +70,22 @@
             "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.9.0/tars/TensorRT-10.9.0.34.Linux.x86_64-gnu.cuda-12.8.tar.gz",
             "strip_prefix": "TensorRT-10.9.0.34",
         },
+        "10.10.0": {
+            "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.10.0/tars/TensorRT-10.10.0.31.Linux.x86_64-gnu.cuda-12.9.tar.gz",
+            "strip_prefix": "TensorRT-10.10.0.31",
+        },
+        "10.11.0": {
+            "urls": "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.11.0/tars/TensorRT-10.11.0.33.Linux.x86_64-gnu.cuda-12.9.tar.gz",
+            "strip_prefix": "TensorRT-10.11.0.33",
+        },
     },
 }
 
 
 def check_new_tensorrt_version(
     major: int = 10, patch_from: int = 0
 ) -> tuple[bool, str, str, str, str]:
+
     def check_file_availability(url: str) -> bool:
         try:
             response = requests.head(url, allow_redirects=True)

MODULE.bazel

@@ -90,10 +90,10 @@ http_archive(
 http_archive(
     name = "torch_l4t",
     build_file = "@//third_party/libtorch:BUILD",
+    sha256 = "6eff643c0a7acda92734cc798338f733ff35c7df1a4434576f5ff7c66fc97319",
     strip_prefix = "torch",
     type = "zip",
     urls = ["https://pypi.jetson-ai-lab.dev/jp6/cu126/+f/6ef/f643c0a7acda9/torch-2.7.0-cp310-cp310-linux_aarch64.whl"],
-    sha256 = "6eff643c0a7acda92734cc798338f733ff35c7df1a4434576f5ff7c66fc97319"
 )
 
 # Download these tarballs manually from the NVIDIA website
@@ -103,18 +103,18 @@ http_archive(
 http_archive(
     name = "tensorrt",
     build_file = "@//third_party/tensorrt/archive:BUILD",
-    strip_prefix = "TensorRT-10.9.0.34",
+    strip_prefix = "TensorRT-10.11.0.33",
     urls = [
-        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.9.0/tars/TensorRT-10.9.0.34.Linux.x86_64-gnu.cuda-12.8.tar.gz",
+        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.11.0/tars/TensorRT-10.11.0.33.Linux.x86_64-gnu.cuda-12.9.tar.gz",
     ],
 )
 
 http_archive(
     name = "tensorrt_sbsa",
     build_file = "@//third_party/tensorrt/archive:BUILD",
-    strip_prefix = "TensorRT-10.9.0.34",
+    strip_prefix = "TensorRT-10.11.0.33",
     urls = [
-        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.9.0/tars/TensorRT-10.9.0.34.Linux.aarch64-gnu.cuda-12.8.tar.gz",
+        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.11.0/tars/TensorRT-10.11.0.33.Linux.aarch64-gnu.cuda-12.9.tar.gz",
     ],
 )
 
@@ -130,9 +130,9 @@ http_archive(
 http_archive(
     name = "tensorrt_win",
     build_file = "@//third_party/tensorrt/archive:BUILD",
-    strip_prefix = "TensorRT-10.9.0.34",
+    strip_prefix = "TensorRT-10.11.0.33",
     urls = [
-        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.9.0/zip/TensorRT-10.9.0.34.Windows.win10.cuda-12.8.zip",
+        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.11.0/zip/TensorRT-10.11.0.33.Windows.win10.cuda-12.9.zip",
     ],
 )
 

README.md

@@ -7,7 +7,7 @@ Torch-TensorRT
 [![Documentation](https://img.shields.io/badge/docs-master-brightgreen)](https://nvidia.github.io/Torch-TensorRT/)
 [![pytorch](https://img.shields.io/badge/PyTorch-2.8-green)](https://download.pytorch.org/whl/nightly/cu128)
 [![cuda](https://img.shields.io/badge/CUDA-12.8-green)](https://developer.nvidia.com/cuda-downloads)
-[![trt](https://img.shields.io/badge/TensorRT-10.9.0-green)](https://github.com/nvidia/tensorrt-llm)
+[![trt](https://img.shields.io/badge/TensorRT-10.11.0-green)](https://github.com/nvidia/tensorrt-llm)
 [![license](https://img.shields.io/badge/license-BSD--3--Clause-blue)](./LICENSE)
 [![Linux x86-64 Nightly Wheels](https://github.com/pytorch/TensorRT/actions/workflows/build-test-linux-x86_64.yml/badge.svg?branch=nightly)](https://github.com/pytorch/TensorRT/actions/workflows/build-test-linux-x86_64.yml)
 [![Linux SBSA Nightly Wheels](https://github.com/pytorch/TensorRT/actions/workflows/build-test-linux-aarch64.yml/badge.svg?branch=nightly)](https://github.com/pytorch/TensorRT/actions/workflows/build-test-linux-aarch64.yml)
@@ -120,8 +120,8 @@ These are the following dependencies used to verify the testcases. Torch-TensorR
 
 - Bazel 8.1.1
 - Libtorch 2.8.0.dev (latest nightly)
-- CUDA 12.8 (CUDA 12.6 on Jetson) 
-- TensorRT 10.9 (TensorRT 10.3 on Jetson)
+- CUDA 12.8 (CUDA 12.6 on Jetson)
+- TensorRT 10.11 (TensorRT 10.3 on Jetson)
 
 ## Deprecation Policy
 

dev_dep_versions.yml

@@ -1,2 +1,2 @@
 __cuda_version__: "12.8"
-__tensorrt_version__: "10.9.0"
+__tensorrt_version__: "10.11.0"

docker/README.md

@@ -15,14 +15,14 @@
 
 ### Instructions
 
-- The example below uses TensorRT 10.9.0.34
+- The example below uses TensorRT 10.11.0.33
 - See <a href="https://github.com/pytorch/TensorRT#dependencies">dependencies</a> for a list of current default dependencies.
 
 > From root of Torch-TensorRT repo
 
 Build:
 ```
-DOCKER_BUILDKIT=1 docker build --build-arg TENSORRT_VERSION=10.9.0 -f docker/Dockerfile -t torch_tensorrt:latest .
+DOCKER_BUILDKIT=1 docker build --build-arg TENSORRT_VERSION=10.11.0 -f docker/Dockerfile -t torch_tensorrt:latest .
 ```
 
 Run:

pyproject.toml

@@ -8,7 +8,7 @@ requires = [
     "cffi>=1.15.1",
     "typing-extensions>=4.7.0",
     "future>=0.18.3",
-    "tensorrt-cu12>=10.9.0,<10.10.0; 'tegra' not in platform_release",
+    "tensorrt-cu12>=10.11.0,<10.12.0; 'tegra' not in platform_release",
     "tensorrt>=10.3.0,<10.4.0; 'tegra' in platform_release",
     "torch>=2.8.0.dev,<2.9.0; 'tegra' not in platform_release",
     "torch>=2.7.0,<2.8.0; 'tegra' in platform_release",
@@ -61,12 +61,12 @@ dependencies = [
     "torch>=2.8.0.dev,<2.9.0; 'tegra' not in platform_release",
     "torch>=2.7.0,<2.8.0; 'tegra' in platform_release",
 
-    "tensorrt>=10.9.0,<10.10.0; 'tegra' not in platform_release",
+    "tensorrt>=10.11.0,<10.12.0; 'tegra' not in platform_release",
     "tensorrt>=10.3.0,<10.4.0; 'tegra' in platform_release",
 
-    "tensorrt-cu12>=10.9.0,<10.10.0; 'tegra' not in platform_release",
-    "tensorrt-cu12-bindings>=10.9.0,<10.10.0; 'tegra' not in platform_release",
-    "tensorrt-cu12-libs>=10.9.0,<10.10.0; 'tegra' not in platform_release",
+    "tensorrt-cu12>=10.11.0,<10.12.0; 'tegra' not in platform_release",
+    "tensorrt-cu12-bindings>=10.11.0,<10.12.0; 'tegra' not in platform_release",
+    "tensorrt-cu12-libs>=10.11.0,<10.12.0; 'tegra' not in platform_release",
 
     "packaging>=23",
 
@@ -101,7 +101,6 @@ torchvision = [
 quantization = ["nvidia-modelopt[all]>=0.27.1"]
 monitoring-tools = ["rich>=13.7.1"]
 jupyter = ["rich[jupyter]>=13.7.1"]
-distributed = ["tensorrt-llm>=0.16.0"]
 
 [project.urls]
 Homepage = "https://pytorch.org/tensorrt"

tests/py/dynamo/runtime/test_000_compilation_settings.py

@@ -1,11 +1,14 @@
 import unittest
+from importlib import metadata
 
 import tensorrt as trt
 import torch
 import torch_tensorrt
 from torch.testing._internal.common_utils import TestCase, run_tests
 from torch_tensorrt.dynamo.utils import is_tegra_platform
 
+from packaging.version import Version
+
 from ..testing_utilities import DECIMALS_OF_AGREEMENT
 
 
@@ -58,7 +61,7 @@ def forward(self, x):
         torch._dynamo.reset()
 
     @unittest.skipIf(
-        is_tegra_platform() and trt._version_ > "10.8",
+        is_tegra_platform() and Version(metadata.version("tensorrt")) > Version("10.8"),
         "DLA is not supported on Jetson platform starting TRT 10.8",
     )
     def test_dla_args(self):

toolchains/ci_workspaces/MODULE.bazel.tmpl

@@ -75,18 +75,18 @@ http_archive = use_repo_rule("@bazel_tools//tools/build_defs/repo:http.bzl", "ht
 http_archive(
     name = "tensorrt",
     build_file = "@//third_party/tensorrt/archive:BUILD",
-    strip_prefix = "TensorRT-10.9.0.34",
+    strip_prefix = "TensorRT-10.11.0.33",
     urls = [
-        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.9.0/tars/TensorRT-10.9.0.34.Linux.x86_64-gnu.cuda-12.8.tar.gz",
+        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.11.0/tars/TensorRT-10.11.0.33.Linux.x86_64-gnu.cuda-12.9.tar.gz",
     ],
 )
 
 http_archive(
     name = "tensorrt_sbsa",
     build_file = "@//third_party/tensorrt/archive:BUILD",
-    strip_prefix = "TensorRT-10.9.0.34",
+    strip_prefix = "TensorRT-10.11.0.33",
     urls = [
-        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.9.0/tars/TensorRT-10.9.0.34.Linux.aarch64-gnu.cuda-12.8.tar.gz",
+        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.11.0/tars/TensorRT-10.11.0.33.Linux.aarch64-gnu.cuda-12.9.tar.gz",
     ],
 )
 
@@ -102,9 +102,9 @@ http_archive(
 http_archive(
     name = "tensorrt_win",
     build_file = "@//third_party/tensorrt/archive:BUILD",
-    strip_prefix = "TensorRT-10.9.0.34",
+    strip_prefix = "TensorRT-10.11.0.33",
     urls = [
-        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.9.0/zip/TensorRT-10.9.0.34.Windows.win10.cuda-12.8.zip",
+        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.11.0/zip/TensorRT-10.11.0.33.Windows.win10.cuda-12.9.zip",
     ],
 )