@@ -108,23 +108,27 @@ On Ubuntu or Debian you can use the command below to install `rapidjson` and `li
108108sudo apt-get install rapidjson-dev libarchive-dev
109109```
110110
111- 2 . Build Python backend
111+ 2 . Build Python backend. Replace \< GIT\_ BRANCH\_ NAME\> with the GitHub branch
112+ that you want to compile. For release branches it should be r\< xx.yy\> (e.g.
113+ r21.06).
112114
113115```
114116$ mkdir build
115117$ cd build
116- $ cmake -DTRITON_ENABLE_GPU=ON -DCMAKE_INSTALL_PREFIX:PATH=`pwd`/install ..
118+ $ cmake -DTRITON_ENABLE_GPU=ON -DTRITON_BACKEND_REPO_TAG=<GIT_BRANCH_NAME> -DTRITON_COMMON_REPO_TAG=<GIT_BRANCH_NAME> -DTRITON_CORE_REPO_TAG=<GIT_BRANCH_NAME> - DCMAKE_INSTALL_PREFIX:PATH=`pwd`/install ..
117119$ make install
118120```
119121
120122The following required Triton repositories will be pulled and used in
121- the build. By default the "main" branch/tag will be used for each repo
122- but the listed CMake argument can be used to override.
123+ the build. If the CMake variables below are not specified, "main" branch
124+ of those repositories will be used. \< GIT\_ BRANCH\_ NAME\> should be the same
125+ as the Python backend repository branch that you are trying to compile.
123126
124- * triton-inference-server/backend: -DTRITON_BACKEND_REPO_TAG=[ tag]
125- * triton-inference-server/common: -DTRITON_COMMON_REPO_TAG=[ tag]
127+ * triton-inference-server/backend: -DTRITON_BACKEND_REPO_TAG=\< GIT\_ BRANCH\_ NAME\>
128+ * triton-inference-server/common: -DTRITON_COMMON_REPO_TAG=\< GIT\_ BRANCH\_ NAME\>
129+ * triton-inference-server/common: -DTRITON_COMMON_REPO_TAG=\< GIT\_ BRANCH\_ NAME\>
126130
127- Set ` DCMAKE_INSTALL_PREFIX ` to the location where the Triton Server is installed. In the released containers,
131+ Set ` - DCMAKE_INSTALL_PREFIX` to the location where the Triton Server is installed. In the released containers,
128132this location is ` /opt/tritonserver ` .
129133
1301343 . Copy example model and configuration
@@ -331,9 +335,12 @@ conda activate python-3-6
331335# NumPy is required for Python models
332336conda install numpy
333337```
334- 3 . Clone the Python backend repository and compile the Python backend stub (replace \< xx.yy\> with the release number):
338+ 3 . Clone the Python backend repository and compile the Python backend stub
339+ (replace \< GIT\_ BRANCH\_ NAME\> with the branch name that you want to use,
340+ for release branches it should be r\< xx.yy\> ):
335341``` bash
336- $ git clone https://github.com/triton-inference-server/python_backend -b < xx.yy>
342+ $ git clone https://github.com/triton-inference-server/python_backend -b
343+ < GIT_BRANCH_NAME>
337344$ cd python_backend
338345$ mkdir build && cd build
339346$ cmake -DTRITON_ENABLE_GPU=ON -DCMAKE_INSTALL_PREFIX:PATH=` pwd` /install ..
0 commit comments