Is this page helpful?

Migrating TensorRT from 10.x to 11.x on NVIDIA DriveOS#

This page describes platform-specific migration considerations for TensorRT on NVIDIA DriveOS when upgrading from TensorRT 10.x to 11.x. For API-level migration details, refer to the C++, Python, trtexec, and Safety Runtime migration pages.

Important

TensorRT 11.x is supported on DriveOS 7.2.5 and later. DriveOS 7.0 customers on DRIVE Orin remain on TensorRT 10.x, which is maintained on a separate branch. If you are running DriveOS 7.0 with TensorRT 10.x, no migration is required.

Platform Eligibility#

Platform

TensorRT Version

Notes

DRIVE Thor (DriveOS 7.2.5+)

11.x

Full upgrade path. Follow the migration steps on this page.

DRIVE Orin (DriveOS 7.0)

10.x (maintained branch)

Hard-forked TensorRT 10.x branch. No upgrade to 11.x.

DRIVE Orin (DriveOS 7.3)

10.x

Out of scope for TensorRT 11.x in this release.

Who Should Migrate#

You need to take action if your DriveOS application does any of the following:

  • Uses implicit INT8 quantization (BuilderFlag::kINT8 with IInt8Calibrator, or --int8 with trtexec).

  • Relies on weakly typed builder flags (BuilderFlag::kFP16, BuilderFlag::kBF16, BuilderFlag::kFP8, or equivalent flags such as --fp16 with trtexec).

  • Uses IPluginV2DynamicExt or IPluginCreator.

  • References removed tactic sources (kCUBLAS, kCUBLAS_LT, kCUDNN).

If your application already uses strong typing, explicit quantization (Q/DQ nodes), and IPluginV3, the upgrade to TensorRT 11.x should require minimal code changes beyond recompilation.

Migration Checklist#

  1. Migrate from weak typing to strong typing. TensorRT 11.x removes all precision-enabling builder flags. Use ModelOpt AutoCast to convert ONNX models to mixed precision before building. Refer to the Strongly Typed Networks section for details.

  2. Migrate from implicit INT8 to explicit quantization. The IInt8Calibrator class and all subclasses have been removed. Use ModelOpt Quantization to produce models with Q/DQ nodes. Refer to the C++ or Python migration page for before/after examples.

  3. Migrate plugins from V2 to V3. IPluginV2DynamicExt and IPluginCreator have been deprecated. All plugins must use IPluginV3 with IPluginCreatorV3One. Refer to the C++ plugin migration section for a complete NonZero plugin example.

  4. Update V2 API calls to V2 replacements. Several methods (getDeviceMemorySize, setWeightStreamingBudget, setDeviceMemory) have been replaced with V2 versions that use int64_t or accept additional parameters. Refer to the V2 API replacements table.

  5. Update safety runtime code (if applicable). Pre-build safety scope validation has been removed. Build failures now surface as tactic errors with layer-level traceback. Refer to the Safety Runtime migration page.

  6. Remove ``–restricted`` from ``trtexec`` scripts. The --restricted flag has been removed in TensorRT 11.x. Remove it from any build scripts or CI pipelines.

  7. Rebuild all engines. Engines built with TensorRT 10.x using the BuilderFlag::kVERSION_COMPATIBLE setting are forward-compatible with the TensorRT 11.x runtime. However, to take advantage of 11.x optimizations and new features, rebuild engines with the 11.x builder.

Version Compatibility on DriveOS#

TensorRT 11.x maintains forward compatibility with TensorRT 10.x engines:

  • Engines built with TensorRT 10.x can run on the TensorRT 11.x runtime.

  • Engines built with TensorRT 8.x or 9.x are not compatible with the TensorRT 11.x runtime. TensorRT 8.x uses CUDA 11.x, which is not supported by TensorRT 11.x.

  • Backward compatibility is maintained within a major release. For example, a version-compatible engine built with TensorRT 11.5 runs on TensorRT 11.5 or any later 11.x release, but not on TensorRT 11.0 through 11.4, and not on TensorRT 10.x.

If your DriveOS deployment currently uses TensorRT 8.x, you must first migrate to TensorRT 10.x using the 8.x to 10.x appendix, then follow this guide for the 10.x to 11.x migration.

CUDA Dependency Changes#

TensorRT 11.x requires CUDA 13.3 or later. Verify that your DriveOS BSP includes a compatible CUDA toolkit before upgrading. TensorRT 11.x drops support for CUDA 11.x, which means TensorRT engines built against CUDA 11.x cannot run on the TensorRT 11.x runtime.

DLA Considerations#

DLA is not supported in TensorRT 11.0; DLA support will be re-introduced in a later minor version update.

Recommendation for Production Deployments#

Tip

For automotive customers already in production on non-safety platforms prior to the TensorRT 11.x release, NVIDIA recommends remaining on the TensorRT 10.x branch to ensure optimal stability and performance. Migrate to 11.x when your next development cycle begins.

For safety customers who must upgrade to 11.x, the TensorRT team provides mitigation plans for all removed APIs and will work directly with customers to ensure a smooth transition. Contact your NVIDIA representative for assistance.