You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 7fe18f0
Browse filesBrowse the repository at this point in the historyBrowse files
Copy file name to clipboardExpand all lines: CHANGELOG.md
+21-3Lines changed: 21 additions & 3 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,15 +1,33 @@
1
1
# TensorRT OSS Release Changelog
2
2
3
+
## 11.5 GA - 2026-11-18
4
+
- General
5
+
- Updated default CUDA version to 13.5.
6
+
7
+
- Parsers
8
+
- Added support for ONNX opset 25 and updated the supported ONNX release to 1.22.0.
9
+
- Updated the CMake build to locate CUDA with `find_package(CUDAToolkit)`; use `-DCUDAToolkit_ROOT` to point at a non-default CUDA install instead of `-DCUDA_TOOLKIT_ROOT_DIR`.
10
+
- Added support for Squeeze when `axis` input is omitted.
11
+
12
+
- Plugins
13
+
- Removed the C++17 fallback for the `PLUGIN_VALIDATE` macros - plugin sources now require C++20.
14
+
15
+
- Samples
16
+
- Added `--independentWeightLayouts` option to `trtexec`.
17
+
- Added `--saveEngineSo` and `--loadEngineSo` options to `trtexec` for safety engines (requires `--safe`).
18
+
- Updated the safety samples to use the CPU-only builder flag.
19
+
20
+
- Tools
21
+
- Bumped Polygraphy version to v0.53.9.
22
+
- Added `model_doc_string` to onnx-graphsurgeon `Graph` to forward the ONNX `ModelProto``doc_string` on export.
23
+
3
24
## 11.4 GA - 2026-10-09
4
25
- Parsers
5
26
- Added `IRefitterObserver` class and `IParser::setRefitObserver` to better handle refittable weights when parsing.
6
27
7
28
- Plugins
8
29
- Added various C++20 updates to plugin source code.
9
30
10
-
- Samples
11
-
- Moved samples/common files only relevant to trtexec to samples/trtexecCommon.
@@ -103,24 +103,24 @@ To build the TensorRT-OSS components, you will first need the following software
103
103
104
104
Else download and extract the TensorRT GA build from [NVIDIA Developer Zone](https://developer.nvidia.com) with the direct links below:
105
105
106
-
-[TensorRT 11.4.0.106 for CUDA 13.4, Linux x86_64](https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/11.4.0/tars/TensorRT-Enterprise-11.4.0.106-Linux-x86_64-cuda-13.4-Release-external.tar.zst)
107
-
-[TensorRT 11.4.0.106 for CUDA 12.9, Linux x86_64](https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/11.4.0/tars/TensorRT-Enterprise-11.4.0.106-Linux-x86_64-cuda-12.9-Release-external.tar.zst)
108
-
-[TensorRT 11.4.0.106 for CUDA 13.4, Windows x86_64](https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/11.4.0/zip/TensorRT-Enterprise-11.4.0.106-Windows-amd64-cuda-13.4-Release-external.zip)
109
-
-[TensorRT 11.4.0.106 for CUDA 12.9, Windows x86_64](https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/11.4.0/zip/TensorRT-Enterprise-11.4.0.106-Windows-amd64-cuda-12.9-Release-external.zip)
106
+
-[TensorRT 11.5.0.63 for CUDA 13.5, Linux x86_64](https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/11.5.0/tars/TensorRT-Enterprise-11.5.0.63-Linux-x86_64-cuda-13.5-Release-external.tar.zst)
107
+
-[TensorRT 11.5.0.63 for CUDA 12.9, Linux x86_64](https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/11.5.0/tars/TensorRT-Enterprise-11.5.0.63-Linux-x86_64-cuda-12.9-Release-external.tar.zst)
108
+
-[TensorRT 11.5.0.63 for CUDA 13.5, Windows x86_64](https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/11.5.0/zip/TensorRT-Enterprise-11.5.0.63-Windows-amd64-cuda-13.5-Release-external.zip)
109
+
-[TensorRT 11.5.0.63 for CUDA 12.9, Windows x86_64](https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/11.5.0/zip/TensorRT-Enterprise-11.5.0.63-Windows-amd64-cuda-12.9-Release-external.zip)
110
110
111
-
**Example: Ubuntu 22.04 on x86-64 with cuda-13.4**
111
+
**Example: Ubuntu 22.04 on x86-64 with cuda-13.5**
112
112
113
113
```bash
114
114
cd~/Downloads
115
-
tar --zstd -xvf TensorRT-Enterprise-11.4.0.106-Linux-x86_64-cuda-13.4-Release-external.tar.zst
116
-
export TRT_LIBPATH=`pwd`/TensorRT-11.4.0.106/lib
115
+
tar --zstd -xvf TensorRT-Enterprise-11.5.0.63-Linux-x86_64-cuda-13.5-Release-external.tar.zst
> NOTE: The default CUDA version used by CMake is 13.4. To override this, for example to 12.9, append `-DCUDA_VERSION=12.9` to the cmake command.
243
+
> NOTE: The default CUDA version used by CMake is 13.5. To override this, for example to 12.9, append `-DCUDA_VERSION=12.9` to the cmake command.
243
244
244
245
- Required CMake build arguments are:
245
246
- `CMAKE_PREFIX_PATH`: Path to the TensorRT package root containing the product CMake config.
@@ -255,6 +256,7 @@ For Linux platforms, we recommend that you generate a docker container for build
255
256
- `BUILD_SAMPLES`: Specify if the samples should be built, for example [`ON`] |`OFF`.
256
257
- `BUILD_SAFE_SAMPLES`: Specify if safety samples should be built, for example [`ON`] |`OFF`.
257
258
- `TRT_SAFETY_INFERENCE_ONLY`: Specify if only build the safety inference components, for example [`ON`] |`OFF`. If turned ON, all other components will be turned OFF except `BUILD_SAFE_SAMPLES`.
259
+
- `TRT_BUILD_PRODUCT`: Select the TensorRT product package to import: `enterprise`, `automotive`, or `safe_inference`. If omitted, the build will infer the product based on the values of `BUILD_SAFE_SAMPLES` and `TRT_SAFETY_INFERENCE_ONLY`.
258
260
- `TRT_BUILD_ENABLE_MULTIDEVICE`: Enable the multi-device sample (`sampleDistCollective`). Use `-DTRT_BUILD_ENABLE_MULTIDEVICE=ON` to build it; requires [NCCL](https://developer.nvidia.com/nccl/nccl-download) >= v2.19, < v3.0.
259
261
- `TRT_BUILD_TESTING`: Build gTests for samples. Requires [gtest](https://github.com/google/googletest) if available; otherwise fetches googletest at configure time.
260
262
@@ -268,6 +270,7 @@ For Linux platforms, we recommend that you generate a docker container for build
@@ -384,6 +389,41 @@ For Linux platforms, we recommend that you generate a docker container for build
384
389
> NOTE: Set `QNX_BASE` to your QNX toolchain installation path.
385
390
> If your CUDA version is not the same as in the example, set`CUDA_VERSION` (for examples that use it in multiple places) or add `-DCUDA_VERSION=<version>` to the cmake command.
386
391
392
+
**Example: Cross-Compile for DOS7 QNX Safety (aarch64)**
393
+
394
+
DOS7 QNX and QNX Safety use the same QNX 8 SDK through `QNX_HOST` and
395
+
`QNX_TARGET`. The QNX Safety build additionally uses SafeCUDA, the matching
396
+
TensorRT SafeInference package, and `cmake_qnx_safe.toolchain`. The standard
397
+
DriveOS QNX Safety build environment provides `PDK_TOP`; the toolchain uses
398
+
the libraries under `$PDK_TOP/drive-qnx-safety/lib-target`.
0 commit comments