TRTorch

Ahead of Time (AOT) compiling for PyTorch JIT

TRTorch is a compiler for PyTorch/TorchScript, targeting NVIDIA GPUs via NVIDIA's TensorRT Deep Learning Optimizer and Runtime. Unlike PyTorch's Just-In-Time (JIT) compiler, TRTorch is an Ahead-of-Time (AOT) compiler, meaning that before you deploy your TorchScript code, you go through an explicit compile step to convert a standard TorchScript program into an module targeting a TensorRT engine. TRTorch operates as a PyTorch extention and compiles modules that integrate into the JIT runtime seamlessly. After compilation using the optimized graph should feel no different than running a TorchScript module. You also have access to TensorRT's suite of configurations at compile time, so you are able to specify operating precision (FP32/FP16/INT8) and other settings for your module.

More Information / System Architecture:

GTC 2020 Talk

Example Usage

C++

#include "torch/script.h"
#include "trtorch/trtorch.h"

...
// Set input datatypes. Allowerd options torch::{kFloat, kHalf, kChar, kInt32, kBool}
// Size of input_dtypes should match number of inputs to the network.
// If input_dtypes is not set, default precision follows traditional PyT / TRT rules
auto input = trtorch::CompileSpec::Input(dims, torch::kHalf)
auto compile_settings = trtorch::CompileSpec({input});
// FP16 execution
compile_settings.enabled_precisions = {torch::kHalf};
// Compile module
auto trt_mod = trtorch::CompileGraph(ts_mod, compile_settings);
// Run like normal
auto results = trt_mod.forward({in_tensor});
// Save module for later
trt_mod.save("trt_torchscript_module.ts");
...

Python

import trtorch

...
compile_settings = {
    "inputs": [trtorch.Input(
        min_shape=[1, 3, 224, 224],
        opt_shape=[1, 3, 512, 512],
        max_shape=[1, 3, 1024, 1024],
        # For static size shape=[1, 3, 224, 224]
        dtype=torch.half, # Datatype of input tensor. Allowed options torch.(float|half|int8|int32|bool)
    )],
    "enabled_precisions": {torch.half}, # Run with FP16
}

trt_ts_module = trtorch.compile(torch_script_module, compile_settings)

input_data = input_data.half()
result = trt_ts_module(input_data)
torch.jit.save(trt_ts_module, "trt_torchscript_module.ts")

Notes on running in lower precisions:

Enabled lower precisions with compile_spec.enabled_precisions

The module should be left in FP32 before compilation (FP16 can support half tensor models)

In FP16 only input tensors by default should be FP16, other precisions use FP32. This can be overrided by setting Input::dtype

Platform Support

Platform	Support
Linux AMD64 / GPU	Supported
Linux aarch64 / GPU	Native Compilation Supported on JetPack-4.4+
Linux aarch64 / DLA	Native Compilation Supported on JetPack-4.4+
Windows / GPU	Unofficial Support
Linux ppc64le / GPU	-

Note: Refer NVIDIA NGC container(https://ngc.nvidia.com/catalog/containers/nvidia:l4t-pytorch) for PyTorch libraries on JetPack.

Dependencies

These are the following dependencies used to verify the testcases. TRTorch can work with other versions, but the tests are not guaranteed to pass.

Bazel 4.0.0
Libtorch 1.9.1 (built with CUDA 11.1)
CUDA 11.1 (10.2 on Jetson)
cuDNN 8.2
TensorRT 8.0.3.4 (TensorRT 8.0.1.6 on Jetson)

Prebuilt Binaries and Wheel files

Releases: https://github.com/NVIDIA/TRTorch/releases

Compiling TRTorch

Installing Dependencies

0. Install Bazel

If you don't have bazel installed, the easiest way is to install bazelisk using the method of you choosing https://github.com/bazelbuild/bazelisk

Otherwise you can use the following instructions to install binaries https://docs.bazel.build/versions/master/install.html

Finally if you need to compile from source (e.g. aarch64 until bazel distributes binaries for the architecture) you can use these instructions

export BAZEL_VERSION=<VERSION>
mkdir bazel
cd bazel
curl -fSsL -O https://github.com/bazelbuild/bazel/releases/download/$BAZEL_VERSION/bazel-$BAZEL_VERSION-dist.zip
unzip bazel-$BAZEL_VERSION-dist.zip
bash ./compile.sh

You need to start by having CUDA installed on the system, LibTorch will automatically be pulled for you by bazel, then you have two options.

1. Building using cuDNN & TensorRT tarball distributions

This is recommended so as to build TRTorch hermetically and insures any bugs are not caused by version issues

Make sure when running TRTorch that these versions of the libraries are prioritized in your $LD_LIBRARY_PATH

You need to download the tarball distributions of TensorRT and cuDNN from the NVIDIA website.
- https://developer.nvidia.com/cudnn
- https://developer.nvidia.com/tensorrt
Place these files in a directory (the directories third_party/dist_dir/[x86_64-linux-gnu | aarch64-linux-gnu] exist for this purpose)
Compile using:

bazel build //:libtrtorch --compilation_mode opt --distdir third_party/dist_dir/[x86_64-linux-gnu | aarch64-linux-gnu]

2. Building using locally installed cuDNN & TensorRT

If you find bugs and you compiled using this method please disclose it in the issue (an ldd dump would be nice too)

Install TensorRT, CUDA and cuDNN on the system before starting to compile.
In WORKSPACE comment out

# Downloaded distributions to use with --distdir
http_archive(
    name = "cudnn",
    urls = ["<URL>",],

    build_file = "@//third_party/cudnn/archive:BUILD",
    sha256 = "<TAR SHA256>",
    strip_prefix = "cuda"
)

http_archive(
    name = "tensorrt",
    urls = ["<URL>",],

    build_file = "@//third_party/tensorrt/archive:BUILD",
    sha256 = "<TAR SHA256>",
    strip_prefix = "TensorRT-<VERSION>"
)

and uncomment

# Locally installed dependencies
new_local_repository(
    name = "cudnn",
    path = "/usr/",
    build_file = "@//third_party/cudnn/local:BUILD"
)

new_local_repository(
   name = "tensorrt",
   path = "/usr/",
   build_file = "@//third_party/tensorrt/local:BUILD"
)

Compile using:

bazel build //:libtrtorch --compilation_mode opt

Debug build

bazel build //:libtrtorch --compilation_mode=dbg

Native compilation on NVIDIA Jetson AGX

bazel build //:libtrtorch --distdir third_party/dist_dir/aarch64-linux-gnu

Note: Please refer installation instructions for Pre-requisites

A tarball with the include files and library can then be found in bazel-bin

Running TRTorch on a JIT Graph

Make sure to add LibTorch to your LD_LIBRARY_PATH
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$(pwd)/bazel-TRTorch/external/libtorch/lib

bazel run //cpp/trtorchexec -- $(realpath <PATH TO GRAPH>) <input-size>

Compiling the Python Package

To compile the python package for your local machine, just run python3 setup.py install in the //py directory. To build wheel files for different python versions, first build the Dockerfile in //py then run the following command

docker run -it -v$(pwd)/..:/workspace/TRTorch build_trtorch_wheel /bin/bash /workspace/TRTorch/py/build_whl.sh

Python compilation expects using the tarball based compilation strategy from above.

How do I add support for a new op...

In TRTorch?

Thanks for wanting to contribute! There are two main ways to handle supporting a new op. Either you can write a converter for the op from scratch and register it in the NodeConverterRegistry or if you can map the op to a set of ops that already have converters you can write a graph rewrite pass which will replace your new op with an equivalent subgraph of supported ops. Its preferred to use graph rewriting because then we do not need to maintain a large library of op converters. Also do look at the various op support trackers in the issues for information on the support status of various operators.

In my application?

The Node Converter Registry is not exposed in the top level API but in the internal headers shipped with the tarball.

You can register a converter for your op using the NodeConverterRegistry inside your application.

Structure of the repo

Component	Description
core	Main JIT ingest, lowering, conversion and runtime implementations
cpp	C++ API and CLI source
examples	Example applications to show different features of TRTorch
py	Python API for TRTorch
tests	Unit tests for TRTorch

Contributing

Take a look at the CONTRIBUTING.md

License

The TRTorch license can be found in the LICENSE file. It is licensed with a BSD Style licence

License

anlongstory/TRTorch

ErrorLooks like something went wrong!

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Languages

Error
Looks like something went wrong!

Packages