Fix examples + halde nccl not found

Author: mreso

cpp/CMakeLists.txt

@@ -40,7 +40,11 @@ list(APPEND CMAKE_PREFIX_PATH ${TORCH_CMAKE_PREFIX_PATH})
 
 find_package(Boost REQUIRED)
 find_package(Torch REQUIRED)
+
 find_library(NCCL_LIBRARY nccl HINTS ${NCCL_PATH})
+if("${NCCL_LIBRARY}" STREQUAL "NOTFOUND")
+  set(NCCL_LIBRARY "")
+endif()
 
 include(FetchContent)
 

cpp/src/utils/CMakeLists.txt

@@ -47,7 +47,7 @@ else()
   target_link_libraries(ts_utils ${CMAKE_DL_LIBS} ${Boost_LIBRARIES} yaml-cpp nlohmann_json fmt)
 endif()
 
-install(TARGETS ts_utils DESTINATION ${torchserve_cpp_SOURCE_DIR}/_build/libs)
+install(TARGETS ts_utils DESTINATION ${CMAKE_INSTALL_PREFIX}/libs)
 
 list(APPEND FOO_SOURCE_FILES ${TS_UTILS_SRC_DIR}/ifoo.hh)
 list(APPEND FOO_SOURCE_FILES ${TS_UTILS_SRC_DIR}/foo.hh)

examples/cpp/aot_inductor/bert/README.md

@@ -7,13 +7,8 @@ Then, this example loads model and runs prediction using libtorch. The handler C
 ### Setup
 1. Follow the instructions in [README.md](../../../../cpp/README.md) to build the TorchServe C++ backend.
 
-```
-cd serve/cpp
-./builld.sh
-```
-
-The build script will create the necessary artifact for this example.
-To recreate these by hand you can follow the prepare_test_files function of the [build.sh](../../../../cpp/build.sh) script.
+The build process will create the necessary artifact for this example.
+To recreate these by hand you can follow the the [CMakeLists.txt](./CMakeLists.txt) file.
 We will need the handler .so file as well as the bert-seq.so and tokenizer.json.
 
 2. Create a [model-config.yaml](model-config.yaml)
@@ -37,7 +32,7 @@ handler:
 ### Generate Model Artifact Folder
 
 ```bash
-torch-model-archiver --model-name bertcppaot --version 1.0 --handler ../../../../cpp/_build/test/resources/examples/aot_inductor/bert_handler/libbert_handler:BertCppHandler --runtime LSP --extra-files index_to_name.json,../../../../cpp/_build/test/resources/examples/aot_inductor/bert_handler/bert-seq.so,../../../../cpp/_build/test/resources/examples/aot_inductor/bert_handler/tokenizer.json  --config-file model-config.yaml --archive-format no-archive
+torch-model-archiver --model-name bertcppaot --version 1.0 --handler ../../../../cpp/build/test/resources/examples/aot_inductor/bert_handler/libbert_handler:BertCppHandler --runtime LSP --extra-files index_to_name.json,../../../../cpp/build/test/resources/examples/aot_inductor/bert_handler/bert-seq.so,../../../../cpp/build/test/resources/examples/aot_inductor/bert_handler/tokenizer.json  --config-file model-config.yaml --archive-format no-archive
 ```
 
 Create model store directory and move the folder `bertcppaot`

examples/cpp/aot_inductor/llama2/README.md

@@ -5,29 +5,24 @@ The handler C++ source code for this examples can be found [here](src/).
 ### Setup
 1. Follow the instructions in [README.md](../../../../cpp/README.md) to build the TorchServe C++ backend.
 
-```
-cd serve/cpp
-./builld.sh
-```
-
-The build script will already create the necessary artifact for this example.
-To recreate these by hand you can follow the prepare_test_files function of the [build.sh](../../../../cpp/build.sh) script.
+The build process will create the necessary artifact for this example.
+To recreate these by hand you can follow the the [CMakeLists.txt](./CMakeLists.txt) file.
 We will need the handler .so file as well as the stories15M.so file containing the model and weights.
 
 2. Copy the handler file
 
 ```bash
 cd ~/serve/examples/cpp/aot_inductor/llama2
-cp ../../../../cpp/_build/test/resources/examples/aot_inductor/llama_handler/libllama_so_handler.so ./
+cp ../../../../cpp/build/test/resources/examples/aot_inductor/llama_handler/libllama_so_handler.so ./
 ```
 We will leave the model .so file in place and just use its absolute path in the next step.
 
 4. Create a [config.json](config.json) with the path of the downloaded model and tokenizer:
 
 ```bash
 echo '{
-"checkpoint_path" : "/home/ubuntu/serve/cpp/_build/test/resources/examples/aot_inductor/llama_handler/stories15M.so",
-"tokenizer_path" : "/home/ubuntu/serve/cpp/_build/test/resources/examples/babyllama/babyllama_handler/tokenizer.bin"
+"checkpoint_path" : "/home/ubuntu/serve/cpp/build/test/resources/examples/aot_inductor/llama_handler/stories15M.so",
+"tokenizer_path" : "/home/ubuntu/serve/cpp/build/test/resources/examples/babyllama/babyllama_handler/tokenizer.bin"
 }' > config.json
 ```
 

examples/cpp/aot_inductor/resnet/README.md

@@ -6,13 +6,8 @@ The handler C++ source code for this examples can be found [here](src).
 ### Setup
 1. Follow the instructions in [README.md](../../../../cpp/README.md) to build the TorchServe C++ backend.
 
-```
-cd serve/cpp
-./builld.sh
-```
-
-The build script will create the necessary artifact for this example.
-To recreate these by hand you can follow the prepare_test_files function of the [build.sh](../../../../cpp/build.sh) script.
+The build process will create the necessary artifact for this example.
+To recreate these by hand you can follow the the [CMakeLists.txt](./CMakeLists.txt) file.
 We will need the handler .so file as well as the resnet50_pt2.so file containing the model and weights.
 
 2. Create a [model-config.yaml](model-config.yaml)
@@ -30,7 +25,7 @@ handler:
 ### Generate Model Artifacts Folder
 
 ```bash
-torch-model-archiver --model-name resnetcppaot --version 1.0 --handler ../../../../cpp/_build/test/resources/examples/aot_inductor/resnet_handler/libresnet_handler:ResnetCppHandler --runtime LSP --extra-files index_to_name.json,../../../../cpp/_build/test/resources/examples/aot_inductor/resnet_handler/resnet50_pt2.so --config-file model-config.yaml --archive-format no-archive
+torch-model-archiver --model-name resnetcppaot --version 1.0 --handler ../../../../cpp/build/test/resources/examples/aot_inductor/resnet_handler/libresnet_handler:ResnetCppHandler --runtime LSP --extra-files index_to_name.json,../../../../cpp/build/test/resources/examples/aot_inductor/resnet_handler/resnet50_pt2.so --config-file model-config.yaml --archive-format no-archive
 ```
 
 Create model store directory and move the folder `resnetcppaot`