Support using onnxruntime 1.16.0 with CUDA 11.4 on Jetson Orin NX (Li…

…nux arm64 GPU). (#1630) * Support using onnxruntime 1.16.0 with CUDA 11.4 on Jetson Orin NX. The pre-built onnxruntime libs are provided by the community using the following command: ```bash ./build.sh --build_shared_lib --config Release --update \ --build --parallel --use_cuda \ --cuda_home /usr/local/cuda \ --cudnn_home /usr/lib/aarch64-linux-gnu 2>&1 | tee my-log.txt ``` See also microsoft/onnxruntime#11226 --- Info about the board: ``` Model: NVIDIA Orin NX T801-16GB - Jetpack 5.1.4 [L4T 35.6.0] ``` ``` nvidia@nvidia-desktop:~/Downloads$ head -n 1 /etc/nv_tegra_release # R35 (release), REVISION: 6.0, GCID: 37391689, BOARD: t186ref, EABI: aarch64, DATE: Wed Aug 28 09:12:27 UTC 2024 nvidia@nvidia-desktop:~/Downloads$ uname -r 5.10.216-tegra nvidia@nvidia-desktop:~/Downloads$ lsb_release -i -r Distributor ID: Ubuntu Release: 20.04 nvidia@nvidia-desktop:~/Downloads$ nvcc -V nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2022 NVIDIA Corporation Built on Wed_Sep_21_10:43:33_PDT_2022 Cuda compilation tools, release 11.8, V11.8.89 Build cuda_11.8.r11.8/compiler.31833905_0 nvidia@nvidia-desktop:~/Downloads$ dpkg -l libcudnn8 Desired=Unknown/Install/Remove/Purge/Hold | Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad) ||/ Name Version Architecture Description +++-==============-====================-============-================================= ii libcudnn8 8.6.0.166-1+cuda11.4 arm64 cuDNN runtime libraries nvidia@nvidia-desktop:~/Downloads$ dpkg -l tensorrt Desired=Unknown/Install/Remove/Purge/Hold | Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad) ||/ Name Version Architecture Description +++-==============-==================-============-================================= ii tensorrt 8.5.2.2-1+cuda11.4 arm64 Meta package for TensorRT ```
k2-fsa · Dec 19, 2024 · 70ee779 · 70ee779
1 parent 5cc60de
commit 70ee779
Show file tree

Hide file tree

Showing 6 changed files with 73 additions and 21 deletions.
diff --git a/.github/workflows/aarch64-linux-gnu-shared.yaml b/.github/workflows/aarch64-linux-gnu-shared.yaml
@@ -34,12 +34,20 @@ concurrency:
 jobs:
   aarch64_linux_gnu_shared:
     runs-on: ${{ matrix.os }}
-    name: aarch64 shared GPU ${{ matrix.gpu }}
+    name: aarch64 shared GPU ${{ matrix.gpu }} ${{ matrix.onnxruntime_version }}
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest]
-        gpu: [ON, OFF]
+        include:
+          - os: ubuntu-latest
+            gpu: ON
+            onnxruntime_version: "1.11.0"
+          - os: ubuntu-latest
+            gpu: ON
+            onnxruntime_version: "1.16.0"
+          - os: ubuntu-latest
+            gpu: OFF
+            onnxruntime_version: ""
 
     steps:
       - uses: actions/checkout@v4
@@ -62,7 +70,7 @@ jobs:
         if: steps.cache-qemu.outputs.cache-hit != 'true'
         run: |
           sudo apt-get update
-          sudo apt-get install autoconf automake autotools-dev ninja-build
+          sudo apt-get install autoconf automake autotools-dev ninja-build libglib2.0-dev.
 
       - name: checkout-qemu
         if: steps.cache-qemu.outputs.cache-hit != 'true'
@@ -159,6 +167,7 @@ jobs:
 
           export BUILD_SHARED_LIBS=ON
           export SHERPA_ONNX_ENABLE_GPU=${{ matrix.gpu }}
+          export SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION=${{ matrix.onnxruntime_version }}
 
           ./build-aarch64-linux-gnu.sh
 
@@ -199,7 +208,7 @@ jobs:
           if [[ ${{ matrix.gpu }} == OFF ]]; then
             dst=${dst}-cpu
           else
-            dst=${dst}-gpu
+            dst=${dst}-gpu-onnxruntime-${{ matrix.onnxruntime_version }}
           fi
           mkdir $dst
 
@@ -223,7 +232,7 @@ jobs:
 
       - uses: actions/upload-artifact@v4
         with:
-          name: sherpa-onnx-linux-aarch64-shared-gpu-${{ matrix.gpu }}
+          name: sherpa-onnx-linux-aarch64-shared-gpu-${{ matrix.gpu }}-onnxruntime-${{ matrix.onnxruntime_version }}
           path: sherpa-onnx-*linux-aarch64-shared*.tar.bz2
 
       # https://huggingface.co/docs/hub/spaces-github-actions

diff --git a/.github/workflows/aarch64-linux-gnu-static.yaml b/.github/workflows/aarch64-linux-gnu-static.yaml
@@ -61,7 +61,7 @@ jobs:
         if: steps.cache-qemu.outputs.cache-hit != 'true'
         run: |
           sudo apt-get update
-          sudo apt-get install autoconf automake autotools-dev ninja-build
+          sudo apt-get install autoconf automake autotools-dev ninja-build libglib2.0-dev.
 
       - name: checkout-qemu
         if: steps.cache-qemu.outputs.cache-hit != 'true'

diff --git a/.github/workflows/arm-linux-gnueabihf.yaml b/.github/workflows/arm-linux-gnueabihf.yaml
@@ -62,7 +62,7 @@ jobs:
         if: steps.cache-qemu.outputs.cache-hit != 'true'
         run: |
           sudo apt-get update
-          sudo apt-get install autoconf automake autotools-dev ninja-build
+          sudo apt-get install autoconf automake autotools-dev ninja-build libglib2.0-dev.
 
       - name: checkout-qemu
         if: steps.cache-qemu.outputs.cache-hit != 'true'

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -46,6 +46,9 @@ option(SHERPA_ONNX_USE_PRE_INSTALLED_ONNXRUNTIME_IF_AVAILABLE "True to use pre-i
 option(SHERPA_ONNX_ENABLE_SANITIZER "Whether to enable ubsan and asan" OFF)
 option(SHERPA_ONNX_BUILD_C_API_EXAMPLES "Whether to enable C API examples" ON)
 
+set(SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION "1.11.0" CACHE STRING "Used only for Linux ARM64 GPU. If you use Jetson nano b01, then please set it to 1.11.0. If you use Jetson Orin NX, then set it to 1.16.0")
+
+
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/lib")
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/lib")
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/bin")

diff --git a/build-aarch64-linux-gnu.sh b/build-aarch64-linux-gnu.sh
@@ -1,4 +1,25 @@
 #!/usr/bin/env bash
+#
+# Usage of this file
+#
+# (1) Build CPU version of sherpa-onnx
+#    ./build-aarch64-linux-gnu.sh
+#
+# (2) Build GPU version of sherpa-onnx
+#
+#   (a) Make sure your board has NVIDIA GPU(s)
+#
+#   (b) For Jetson Nano B01 (using CUDA 10.2)
+#
+#       export SHERPA_ONNX_ENABLE_GPU=ON
+#       export SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION=1.11.0
+#       ./build-aarch64-linux-gnu.sh
+#
+#   (c) For Jetson Orin NX (using CUDA 11.4)
+#
+#       export SHERPA_ONNX_ENABLE_GPU=ON
+#       export SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION=1.16.0
+#       ./build-aarch64-linux-gnu.sh
 
 if command -v aarch64-none-linux-gnu-gcc  &> /dev/null; then
   ln -svf $(which aarch64-none-linux-gnu-gcc) ./aarch64-linux-gnu-gcc
@@ -47,18 +68,18 @@ fi
 if [[ x"$SHERPA_ONNX_ENABLE_GPU" == x"" ]]; then
   # By default, use CPU
   SHERPA_ONNX_ENABLE_GPU=OFF
-
-  # If you use GPU, then please make sure you have NVIDIA GPUs on your board.
-  # It uses onnxruntime 1.11.0.
-  #
-  # Tested on Jetson Nano B01
 fi
 
 if [[ x"$SHERPA_ONNX_ENABLE_GPU" == x"ON" ]]; then
   # Build shared libs if building GPU is enabled.
   BUILD_SHARED_LIBS=ON
 fi
 
+if [[ x"$SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION" == x"" ]]; then
+  # Used only when SHERPA_ONNX_ENABLE_GPU is ON
+  SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION="1.11.0"
+fi
+
 cmake \
   -DBUILD_PIPER_PHONMIZE_EXE=OFF \
   -DBUILD_PIPER_PHONMIZE_TESTS=OFF \
@@ -75,6 +96,7 @@ cmake \
   -DSHERPA_ONNX_ENABLE_JNI=OFF \
   -DSHERPA_ONNX_ENABLE_C_API=ON \
   -DSHERPA_ONNX_ENABLE_WEBSOCKET=ON \
+  -DSHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION=$SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION \
   -DCMAKE_TOOLCHAIN_FILE=../toolchains/aarch64-linux-gnu.toolchain.cmake \
   ..
 

diff --git a/cmake/onnxruntime-linux-aarch64-gpu.cmake b/cmake/onnxruntime-linux-aarch64-gpu.cmake
@@ -18,19 +18,37 @@ if(NOT SHERPA_ONNX_ENABLE_GPU)
   message(FATAL_ERROR "This file is for NVIDIA GPU only. Given SHERPA_ONNX_ENABLE_GPU: ${SHERPA_ONNX_ENABLE_GPU}")
 endif()
 
-set(onnxruntime_URL  "https://github.com/csukuangfj/onnxruntime-libs/releases/download/v1.11.0/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2")
-set(onnxruntime_URL2 "https://hf-mirror.com/csukuangfj/onnxruntime-libs/resolve/main/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2")
-set(onnxruntime_HASH "SHA256=36eded935551e23aead09d4173bdf0bd1e7b01fdec15d77f97d6e34029aa60d7")
+message(WARNING "\
+SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION: ${SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION}
+If you use Jetson nano b01, then please pass
+   -DSHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION=1.11.0
+to cmake (You need to make sure CUDA 10.2 is available on your board).
+
+If you use Jetson Orin NX, then please pass
+   -DSHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION=1.16.0
+to cmake (You need to make sure CUDA 11.4 is available on your board).
+")
+
+set(v ${SHERPA_ONNX_LINUX_ARM64_GPU_ONNXRUNTIME_VERSION})
+
+set(onnxruntime_URL  "https://github.com/csukuangfj/onnxruntime-libs/releases/download/v${v}/onnxruntime-linux-aarch64-gpu-${v}.tar.bz2")
+set(onnxruntime_URL2 "https://hf-mirror.com/csukuangfj/onnxruntime-libs/resolve/main/onnxruntime-linux-aarch64-gpu-${v}.tar.bz2")
+
+if(v STREQUAL "1.11.0")
+  set(onnxruntime_HASH "SHA256=36eded935551e23aead09d4173bdf0bd1e7b01fdec15d77f97d6e34029aa60d7")
+else()
+  set(onnxruntime_HASH "SHA256=4c09d5acf2c2682b4eab1dc2f1ad98fc1fde5f5f1960063e337983ba59379a4b")
+endif()
 
 # If you don't have access to the Internet,
 # please download onnxruntime to one of the following locations.
 # You can add more if you want.
 set(possible_file_locations
-  $ENV{HOME}/Downloads/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
-  ${CMAKE_SOURCE_DIR}/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
-  ${CMAKE_BINARY_DIR}/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
-  /tmp/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
-  /star-fj/fangjun/download/github/onnxruntime-linux-aarch64-gpu-1.11.0.tar.bz2
+  $ENV{HOME}/Downloads/onnxruntime-linux-aarch64-gpu-${v}.tar.bz2
+  ${CMAKE_SOURCE_DIR}/onnxruntime-linux-aarch64-gpu-${v}.tar.bz2
+  ${CMAKE_BINARY_DIR}/onnxruntime-linux-aarch64-gpu-${v}.tar.bz2
+  /tmp/onnxruntime-linux-aarch64-gpu-${v}.tar.bz2
+  /star-fj/fangjun/download/github/onnxruntime-linux-aarch64-gpu-${v}.tar.bz2
 )
 
 foreach(f IN LISTS possible_file_locations)