fix: Upgrade release/1.4 to Torch 2.0.1 + TensorRT 8.6.1 (#1896)

Author: gs-olive

.circleci/config.yml

@@ -119,7 +119,7 @@ These are the following dependencies used to verify the testcases. Torch-TensorR
 - Libtorch 2.0.1 (built with CUDA 11.8)
 - CUDA 11.8
 - cuDNN 8.8.0
-- TensorRT 8.6.0
+- TensorRT 8.6.1
 
 ## Prebuilt Binaries and Wheel files
 

README.md

@@ -51,17 +51,17 @@ new_local_repository(
 http_archive(
     name = "libtorch",
     build_file = "@//third_party/libtorch:BUILD",
-    sha256 = "c5174f18c0866421a5738d389aaea0c02f32a1a5be5f0747dc8dd0d96034c9b0",
+    sha256 = "843ad19e769a189758fd6a21bfced9024494b52344f4bc4fb75f75d36e6ea0c7",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/test/cu118/libtorch-cxx11-abi-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/cu118/libtorch-cxx11-abi-shared-with-deps-2.0.1%2Bcu118.zip"],
 )
 
 http_archive(
     name = "libtorch_pre_cxx11_abi",
     build_file = "@//third_party/libtorch:BUILD",
-    sha256 = "cc19b398cf435e0e34f347ef90fc11c2a42703998330a9c4a9fb0d2291737df7",
+    sha256 = "668a5816ee588e5d96be4803cbbd34263314197cbbb320f43ad25a8c626c02e1",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/test/cu118/libtorch-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/cu118/libtorch-shared-with-deps-2.0.1%2Bcu118.zip"],
 )
 
 # Download these tarballs manually from the NVIDIA website
@@ -81,10 +81,10 @@ http_archive(
 http_archive(
     name = "tensorrt",
     build_file = "@//third_party/tensorrt/archive:BUILD",
-    sha256 = "c1732a1093c57ab79fa0b687f061be369e449c9c17792b660f3663ecd8fa7b63",
-    strip_prefix = "TensorRT-8.6.0.12",
+    sha256 = "15bfe6053d45feec45ecc7123a9106076b0b43fa0435f242d89dca0778337759",
+    strip_prefix = "TensorRT-8.6.1.6",
     urls = [
-        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/secure/8.6.0/tars/TensorRT-8.6.0.12.Linux.x86_64-gnu.cuda-11.8.tar.gz",
+        "https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/secure/8.6.1/tars/TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-11.8.tar.gz",
     ],
 )
 

WORKSPACE

@@ -1,10 +1,13 @@
 # Base image starts with CUDA
 ARG BASE_IMG=nvidia/cuda:11.8.0-devel-ubuntu22.04
 FROM ${BASE_IMG} as base
+ENV BASE_IMG=nvidia/cuda:11.8.0-devel-ubuntu22.04
 
 ARG TENSORRT_VERSION
+ENV TENSORRT_VERSION=${TENSORRT_VERSION}
 RUN test -n "$TENSORRT_VERSION" || (echo "No tensorrt version specified, please use --build-arg TENSORRT_VERSION=x.y.z to specify a version." && exit 1)
 ARG CUDNN_VERSION
+ENV CUDNN_VERSION=${CUDNN_VERSION}
 RUN test -n "$CUDNN_VERSION" || (echo "No cudnn version specified, please use --build-arg CUDNN_VERSION=x.y.z to specify a version." && exit 1)
 
 ARG PYTHON_VERSION=3.10
@@ -44,7 +47,7 @@ RUN apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/
 RUN add-apt-repository "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/ /"
 RUN apt-get update
 
-RUN apt-get install -y libnvinfer8=${TENSORRT_VERSION}* libnvinfer-plugin8=${TENSORRT_VERSION}* libnvinfer-dev=${TENSORRT_VERSION}* libnvinfer-plugin-dev=${TENSORRT_VERSION}* libnvonnxparsers8=${TENSORRT_VERSION}-1* libnvonnxparsers-dev=${TENSORRT_VERSION}-1* libnvparsers8=${TENSORRT_VERSION}-1*  libnvparsers-dev=${TENSORRT_VERSION}-1*
+RUN apt-get install -y libnvinfer8=${TENSORRT_VERSION}.* libnvinfer-plugin8=${TENSORRT_VERSION}.* libnvinfer-dev=${TENSORRT_VERSION}.* libnvinfer-plugin-dev=${TENSORRT_VERSION}.* libnvonnxparsers8=${TENSORRT_VERSION}.* libnvonnxparsers-dev=${TENSORRT_VERSION}.* libnvparsers8=${TENSORRT_VERSION}.*  libnvparsers-dev=${TENSORRT_VERSION}.*
 
 # Setup Bazel via Bazelisk
 RUN wget -q https://github.com/bazelbuild/bazelisk/releases/download/v1.16.0/bazelisk-linux-amd64 -O /usr/bin/bazel &&\
@@ -71,7 +74,18 @@ WORKDIR /workspace/torch_tensorrt/src
 RUN cp ./docker/WORKSPACE.docker WORKSPACE
 
 # Symlink the path pyenv is using for python with the /opt directory for package sourcing
-RUN ln -s "`pyenv which python | xargs dirname | xargs dirname`/lib/python$PYTHON_VERSION/site-packages" "/opt/python3"
+RUN mkdir -p "/opt/python3/" &&\
+    ln -s "`pyenv which python | xargs dirname | xargs dirname`/lib/python$PYTHON_VERSION/site-packages" "/opt/python3/"
+
+# Extract base image cuda version (everything after :, before -, before final ., in BASE_IMG)
+# Ensure the default cuda folder agrees with the version in the base image
+RUN CUDA_BASE_IMG_VERSION_INTERMEDIATE=`echo ${BASE_IMG#*:}` &&\
+    CUDA_BASE_IMG_VERSION=`echo ${CUDA_BASE_IMG_VERSION_INTERMEDIATE%%-*}` &&\
+    CUDA_MAJOR_MINOR_VERSION=`echo ${CUDA_BASE_IMG_VERSION%.*}` &&\
+    rm -fr /usr/local/cuda &&\
+    ln -s /usr/local/cuda-${CUDA_MAJOR_MINOR_VERSION} /usr/local/cuda
+
+ENV CUDA_HOME=/usr/local/cuda
 
 # This script builds both libtorchtrt bin/lib/include tarball and the Python wheel, in dist/
 RUN bash ./docker/dist-build.sh

docker/Dockerfile

@@ -3,7 +3,7 @@
 * Use `Dockerfile` to build a container which provides the exact development environment that our master branch is usually tested against.
 
 * The `Dockerfile` currently uses <a href="https://github.com/bazelbuild/bazelisk">Bazelisk</a> to select the Bazel version, and uses the exact library versions of Torch and CUDA listed in <a href="https://github.com/pytorch/TensorRT#dependencies">dependencies</a>.
-  * The desired versions of CUDNN and TensorRT must be specified as build-args, with major, minor, and patch versions as in: `--build-arg TENSORRT_VERSION=a.b.c --build-arg CUDNN_VERSION=x.y.z`
+  * The desired versions of CUDNN and TensorRT must be specified as build-args, with major and minor versions as in: `--build-arg TENSORRT_VERSION=a.b --build-arg CUDNN_VERSION=x.y`
   * [**Optional**] The desired base image be changed by explicitly setting a base image, as in `--build-arg BASE_IMG=nvidia/cuda:11.8.0-devel-ubuntu22.04`, though this is optional
   * [**Optional**] Additionally, the desired Python version can be changed by explicitly setting a version, as in `--build-arg PYTHON_VERSION=3.10`, though this is optional as well.
 
@@ -17,14 +17,14 @@ Note: By default the container uses the `pre-cxx11-abi` version of Torch + Torch
 
 ### Instructions
 
-- The example below uses CUDNN 8.8.0 and TensorRT 8.6.0
+- The example below uses CUDNN 8.8 and TensorRT 8.6
 - 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=8.6.0 --build-arg CUDNN_VERSION=8.8.0 -f docker/Dockerfile -t torch_tensorrt:latest .
+DOCKER_BUILDKIT=1 docker build --build-arg TENSORRT_VERSION=8.6 --build-arg CUDNN_VERSION=8.8 -f docker/Dockerfile -t torch_tensorrt:latest .
 ```
 
 Run:

docker/README.md

@@ -3,17 +3,15 @@
 TOP_DIR=$(cd $(dirname $0); pwd)/..
 
 if [[ -z "${USE_CXX11}" ]]; then
-    BUILD_CMD="python3 setup.py bdist_wheel"
+    BUILD_CMD="python setup.py bdist_wheel"
 else
-    BUILD_CMD="python3 setup.py bdist_wheel  --use-cxx11-abi"
+    BUILD_CMD="python setup.py bdist_wheel  --use-cxx11-abi"
 fi
 
 cd ${TOP_DIR} \
     && mkdir -p dist && cd py \
-    && pip install -r requirements.txt
-
-# Symlink the path pyenv is using for python with the /opt directory for package sourcing
-ln -s "`pyenv which python | xargs dirname | xargs dirname`/lib/python$PYTHON_VERSION/site-packages" "/opt/python3"
+    && pip install -r requirements.txt \
+    && pip install wheel
 
 # Build Torch-TRT
 MAX_JOBS=1 LANG=en_US.UTF-8 LANGUAGE=en_US:en LC_ALL=en_US.UTF-8 ${BUILD_CMD} $* || exit 1

docker/dist-build.sh

@@ -1,8 +1,8 @@
 numpy
 packaging
 pybind11==2.6.2
---extra-index-url https://download.pytorch.org/whl/test/cu118
+--extra-index-url https://download.pytorch.org/whl/cu118
 torch==2.0.1
 torchvision==0.15.2
 --extra-index-url https://pypi.ngc.nvidia.com
-tensorrt==8.6.0
+tensorrt==8.6.1

py/requirements.txt

@@ -12,7 +12,7 @@ FX2TRT is merged as FX module in Torch-TensorRT
     $ conda install pytorch torchvision torchtext cudatoolkit=11.8 -c pytorch-nightly
     # Install TensorRT python package
     $ pip3 install nvidia-pyindex
-    $ pip3 install tensorrt==8.6.0.12
+    $ pip3 install tensorrt==8.6.1
     $ git clone https://github.com/pytorch/TensorRT.git
     $ cd TensorRT/py && python setup.py install --fx-only && cd ..
     $ python -c "import torch_tensorrt.fx"

py/torch_tensorrt/fx/README.md

@@ -1,4 +1,4 @@
-__version__ = "1.4.0.rc0"
+__version__ = "1.4.0"
 __cuda_version__ = "11.8"
 __cudnn_version__ = "8.8"
 __tensorrt_version__ = "8.6"

py/versions.py

@@ -56,17 +56,17 @@ new_local_repository(
 http_archive(
     name = "libtorch",
     build_file = "@//third_party/libtorch:BUILD",
-    sha256 = "c5174f18c0866421a5738d389aaea0c02f32a1a5be5f0747dc8dd0d96034c9b0",
+    sha256 = "843ad19e769a189758fd6a21bfced9024494b52344f4bc4fb75f75d36e6ea0c7",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/test/cu118/libtorch-cxx11-abi-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/cu118/libtorch-cxx11-abi-shared-with-deps-2.0.1%2Bcu118.zip"],
 )
 
 http_archive(
     name = "libtorch_pre_cxx11_abi",
     build_file = "@//third_party/libtorch:BUILD",
-    sha256 = "cc19b398cf435e0e34f347ef90fc11c2a42703998330a9c4a9fb0d2291737df7",
+    sha256 = "668a5816ee588e5d96be4803cbbd34263314197cbbb320f43ad25a8c626c02e1",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/test/cu118/libtorch-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/cu118/libtorch-shared-with-deps-2.0.1%2Bcu118.zip"],
 )
 
 ####################################################################################

toolchains/ci_workspaces/WORKSPACE.x86_64.release.rhel

@@ -56,17 +56,17 @@ new_local_repository(
 http_archive(
     name = "libtorch",
     build_file = "@//third_party/libtorch:BUILD",
-    sha256 = "c5174f18c0866421a5738d389aaea0c02f32a1a5be5f0747dc8dd0d96034c9b0",
+    sha256 = "843ad19e769a189758fd6a21bfced9024494b52344f4bc4fb75f75d36e6ea0c7",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/test/cu118/libtorch-cxx11-abi-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/cu118/libtorch-cxx11-abi-shared-with-deps-2.0.1%2Bcu118.zip"],
 )
 
 http_archive(
     name = "libtorch_pre_cxx11_abi",
     build_file = "@//third_party/libtorch:BUILD",
-    sha256 = "cc19b398cf435e0e34f347ef90fc11c2a42703998330a9c4a9fb0d2291737df7",
+    sha256 = "668a5816ee588e5d96be4803cbbd34263314197cbbb320f43ad25a8c626c02e1",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/test/cu118/libtorch-shared-with-deps-latest.zip"],
+    urls = ["https://download.pytorch.org/libtorch/cu118/libtorch-shared-with-deps-2.0.1%2Bcu118.zip"],
 )
 
 ####################################################################################