Bump dependencies for 23.11 (#369)

* Bump versions for release 23.11 * Update Treelite to 3.9.1
triton-inference-server · Nov 4, 2023 · cb2ec6a · cb2ec6a
1 parent b0b5482
commit cb2ec6a
Show file tree

Hide file tree

Showing 11 changed files with 31 additions and 31 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -24,8 +24,8 @@ set(TRITON_CORE_REPO_TAG "main" CACHE STRING "Tag for triton-inference-server/co
 set(TRITON_BACKEND_REPO_TAG "main" CACHE STRING "Tag for triton-inference-server/backend repo")
 
 if(TRITON_FIL_DOCKER_BUILD)
-  set(TRITON_BUILD_CONTAINER "nvcr.io/nvidia/tritonserver:23.08-py3" CACHE STRING "Build image for Dockerized builds")
-  set(TRITON_BUILD_CONTAINER_VERSION "23.08" CACHE STRING "Triton version for Dockerized builds")
+  set(TRITON_BUILD_CONTAINER "nvcr.io/nvidia/tritonserver:23.09-py3" CACHE STRING "Build image for Dockerized builds")
+  set(TRITON_BUILD_CONTAINER_VERSION "23.09" CACHE STRING "Triton version for Dockerized builds")
 
   add_custom_command(
     OUTPUT fil/libtriton_fil.so $<$<BOOL:${TRITON_ENABLE_GPU}>:fil/libcuml++.so>
@@ -70,7 +70,7 @@ else()
   ##############################################################################
   # - Prepare rapids-cmake -----------------------------------------------------
   file(DOWNLOAD
-    https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-23.08/RAPIDS.cmake
+    https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-23.10/RAPIDS.cmake
       ${CMAKE_BINARY_DIR}/RAPIDS.cmake)
   include(${CMAKE_BINARY_DIR}/RAPIDS.cmake)
   include(rapids-cmake)
@@ -89,7 +89,7 @@ else()
   option(DISABLE_DEPRECATION_WARNINGS "Disable depreaction warnings " ON)
   option(NVTX "Enable nvtx markers" OFF)
   set(BACKEND_FOLDER "/opt/tritonserver/backends" CACHE STRING "Triton backend folder path")
-  set(RAPIDS_DEPENDENCIES_VERSION "23.08" CACHE STRING "RAPIDS projects dependencies version")
+  set(RAPIDS_DEPENDENCIES_VERSION "23.10" CACHE STRING "RAPIDS projects dependencies version")
   option(TRITON_FIL_USE_TREELITE_STATIC "Link Treelite statically in libtriton_fil.so and cuml++.so" ON)
 
 

diff --git a/README.md b/README.md
@@ -152,7 +152,7 @@ dynamic_batching {}
 ```
 docker run -p 8000:8000 -p 8001:8001 --gpus all \
   -v ${PWD}/model_repository:/models \
-  nvcr.io/nvidia/tritonserver:23.08-py3 \
+  nvcr.io/nvidia/tritonserver:23.09-py3 \
   tritonserver --model-repository=/models
 ```
 

diff --git a/build.sh b/build.sh
@@ -168,7 +168,7 @@ DOCKER_ARGS="$DOCKER_ARGS --build-arg TRITON_ENABLE_GPU=${TRITON_ENABLE_GPU}"
 
 if [ -z $RAPIDS_VERSION ]
 then
-  RAPIDS_VERSION=23.08
+  RAPIDS_VERSION=23.10
 else
   DOCKER_ARGS="$DOCKER_ARGS --build-arg RAPIDS_DEPENDENCIES_VERSION=${RAPIDS_VERSION}"
 fi
@@ -190,7 +190,7 @@ if [ -z $TRITON_VERSION ] && [ $HOST_BUILD -eq 1 ]
 then
   # Must use a version compatible with a released backend image in order to
   # test a host build, so default to latest release branch rather than main
-  TRITON_VERSION=23.08
+  TRITON_VERSION=23.09
 fi
 
 if [ ! -z $TRITON_VERSION ]

diff --git a/cmake/thirdparty/get_cuml.cmake b/cmake/thirdparty/get_cuml.cmake
@@ -57,6 +57,6 @@ endfunction()
 # CPM_raft_SOURCE=/path/to/local/raft
 find_and_configure_cuml(VERSION    ${RAPIDS_TRITON_MIN_VERSION_rapids_projects}
                         FORK       rapidsai
-                        PINNED_TAG branch-23.08
+                        PINNED_TAG branch-23.10
                         USE_TREELITE_STATIC ${TRITON_FIL_USE_TREELITE_STATIC}
                         )
diff --git a/cmake/thirdparty/get_treelite.cmake b/cmake/thirdparty/get_treelite.cmake
@@ -35,7 +35,7 @@ function(find_and_configure_treelite)
             GIT_TAG          ${PKG_PINNED_TAG}
             OPTIONS
               "USE_OPENMP ON"
-              "BUILD_STATIC_LIBS ${PKG_BUILD_STATIC_LIBS}"
+              "Treelite_BUILD_STATIC_LIBS ${PKG_BUILD_STATIC_LIBS}"
     )
 
 
@@ -90,6 +90,6 @@ function(find_and_configure_treelite)
     rapids_export_find_package_root(BUILD Treelite [=[${CMAKE_CURRENT_LIST_DIR}]=] cuml-exports)
 endfunction()
 
-find_and_configure_treelite(VERSION     3.2.0
-                        PINNED_TAG  ee697b3b58d5f51623dd3b308f290581b58dbe5d
+find_and_configure_treelite(VERSION     3.9.1
+                        PINNED_TAG  346d92547295417676f499ce2dd4fff946b9042a
                         BUILD_STATIC_LIBS ${TRITON_FIL_USE_TREELITE_STATIC})
diff --git a/conda/environments/triton_benchmark.yml b/conda/environments/triton_benchmark.yml
@@ -5,7 +5,7 @@ channels:
   - rapidsai
 dependencies:
   - cuda-version=11.8
-  - cudf=23.08
+  - cudf=23.10
   - libcusolver<=11.4.1.48
   - libcusparse<=12.0
   - matplotlib
@@ -15,4 +15,4 @@ dependencies:
   - pip:
       - tritonclient[all]
       - protobuf==3.20.1
-      - git+https://github.com/rapidsai/rapids-triton.git@branch-23.08#subdirectory=python
+      - git+https://github.com/rapidsai/rapids-triton.git@branch-23.10#subdirectory=python
diff --git a/conda/environments/triton_test.yml b/conda/environments/triton_test.yml
@@ -7,8 +7,8 @@ dependencies:
   - aws-sdk-cpp
   - clang-tools=11.1.0
   - cuda-version=11.8
-  - cudf=23.08
-  - cuml=23.08
+  - cudf=23.10
+  - cuml=23.10
   - flake8
   - hypothesis<6.46.8
   - lightgbm
@@ -22,4 +22,4 @@ dependencies:
   - pip:
       - tritonclient[all]
       - protobuf==3.20.1
-      - git+https://github.com/rapidsai/rapids-triton.git@branch-23.08#subdirectory=python
+      - git+https://github.com/rapidsai/rapids-triton.git@branch-23.10#subdirectory=python
diff --git a/conda/environments/triton_test_no_client.yml b/conda/environments/triton_test_no_client.yml
@@ -7,8 +7,8 @@ dependencies:
   - aws-sdk-cpp
   - clang-tools=11.1.0
   - cuda-version=11.8
-  - cudf=23.08
-  - cuml=23.08
+  - cudf=23.10
+  - cuml=23.10
   - flake8
   - hypothesis<6.46.8
   - lightgbm

diff --git a/ops/Dockerfile b/ops/Dockerfile
@@ -3,7 +3,7 @@
 # Arguments for controlling build details
 ###########################################################################################
 # Version of Triton to use
-ARG TRITON_VERSION=23.08
+ARG TRITON_VERSION=23.10
 # Base container image
 ARG BASE_IMAGE=nvcr.io/nvidia/tritonserver:${TRITON_VERSION}-py3
 # Whether or not to enable GPU build
@@ -135,7 +135,7 @@ ENV TRITON_ENABLE_GPU=$TRITON_ENABLE_GPU
 ARG TRITON_ENABLE_STATS=ON
 ENV TRITON_ENABLE_GPU=$TRITON_ENABLE_GPU
 
-ARG RAPIDS_DEPENDENCIES_VERSION=23.08
+ARG RAPIDS_DEPENDENCIES_VERSION=23.10
 ENV RAPIDS_DEPENDENCIES_VERSION=$RAPIDS_DEPENDENCIES_VERSION
 
 ARG TRITON_FIL_USE_TREELITE_STATIC=ON

diff --git a/scripts/environment.yml b/scripts/environment.yml
@@ -5,7 +5,7 @@ channels:
   - rapidsai
 dependencies:
   - cuda-version=11.8
-  - cuml=23.08
+  - cuml=23.10
   - python
   - scikit-learn
   - treelite
diff --git a/src/tl_utils.h b/src/tl_utils.h
@@ -25,6 +25,7 @@
 #include <cstddef>
 #include <cstdio>
 #include <filesystem>
+#include <fstream>
 #include <rapids_triton/exceptions.hpp>
 #include <sstream>
 
@@ -54,21 +55,20 @@ load_tl_base_model(
         result = treelite::frontend::LoadLightGBMModel(model_file.c_str());
         break;
       case SerializationFormat::treelite: {
-        auto* file = std::fopen(model_file.c_str(), "rb");
-        if (file == nullptr) {
-          auto log_stream = std::stringstream{};
-          log_stream << "Could not open model file " << model_file;
-          throw rapids::TritonException(
-              rapids::Error::Unavailable, log_stream.str());
-        }
+        auto file = std::fstream{model_file.c_str()};
         try {
-          result = treelite::Model::DeserializeFromFile(file);
+          if (file.is_open()) {
+            result = treelite::Model::DeserializeFromStream(file);
+          } else {
+            auto log_stream = std::stringstream{};
+            log_stream << "Could not open model file " << model_file;
+            throw rapids::TritonException(
+                rapids::Error::Unavailable, log_stream.str());
+          }
         }
         catch (treelite::Error const& err) {
-          std::fclose(file);
           throw;
         }
-        std::fclose(file);
         break;
       }
     }