### Install Protobuf for Operator Compilation

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Ensures the correct protobuf version (<=3.20.x) is installed for operator compilation. This command installs version 3.20.0, which is critical for compatibility. Use `--user` for non-root installations.

```bash
pip3 install protobuf==3.20.0
# For non-root users:
pip3 install protobuf==3.20.0 --user
```

--------------------------------

### Custom Operator Package Installation Parameters

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Details the command-line arguments for installing the custom operator .run package. Supports quiet installation and specifying a custom installation path.

```APIDOC
Custom Operator Package Installation:

Usage: ./CANN-custom_ops-<cann_version>-linux.<arch>.run [OPTIONS]

Options:
  --quiet
    Description: Performs a quiet installation without user prompts.

  --install-path=<path>
    Description: Specifies a custom directory to install the operator package.
    Example: ./build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run --quiet --install-path=/opt/my_operators

  --help
    Description: Displays help information for the installation command.
```

--------------------------------

### Install CANN Development Toolkit (Default Path)

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Command to install the CANN development toolkit and operator binary packages using default installation paths. The toolkit is installed in `/usr/local/Ascend/ascend-toolkit/latest` for root users or `$HOME/Ascend/ascend-toolkit/latest` for non-root users.

```bash
# CANN development toolkit installation:
./Ascend-cann-toolkit_<cann_version>_linux-<arch>.run --install
# Operator binary package installation:
./Ascend-cann-kernels-<soc_version>_<cann_version>_linux.run --install
```

--------------------------------

### AclNN Interface Verification Example Structure

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Illustrates the directory structure and key files for verifying operator functionality using the AclNN interface. Includes scripts for data generation, compilation, execution, and result verification.

```plain text
AclNNInvocationNaive
    ├── CMakeLists.txt
    ├── gen_data.py
    ├── main.cpp
    ├── README.md
    ├── run.sh
    └── verify_result.py
```

--------------------------------

### Install CANN Development Toolkit (Custom Path)

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Command to install the CANN development toolkit and operator binary packages to a specified installation path. This allows flexibility in choosing the installation directory for the toolkit and its associated kernel packages.

```bash
# CANN development toolkit installation:
./Ascend-cann-toolkit_<cann_version>_linux-<arch>.run --install --install-path=${install_path}
# Operator binary package installation:
./Ascend-cann-kernels-<soc_version>_<cann_version>_linux.run --install --install-path=${install_path}
```

--------------------------------

### Compile and Install Googletest

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Provides instructions to compile and install the googletest framework from source, specifically checking out the release-1.11.0 tag. This is required for running Unit Tests (UT). The CMake command includes flags for ABI compatibility.

```bash
git checkout release-1.11.0
mkdir temp && cd temp
cmake .. -DCMAKE_CXX_FLAGS="-fPIC -D_GLIBCXX_USE_CXX11_ABI=0"
make
make install
# For non-root users:
# sudo make install
```

--------------------------------

### ST Test Verification Example Structure

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Shows the typical directory structure for System Test (ST) verification of custom operators. It includes JSON test case definitions, configuration files, and optional benchmark scripts.

```plain text
st
  ├── AddCustom_case_all_type.json
  ├── msopst.ini
  ├── README.md
  └── test_add_custom.py
```

--------------------------------

### Compile and Install nlohmann_json Library

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Details the process for compiling and installing the nlohmann_json library from source, targeting version v3.11.3. This is recommended for JSON handling. Tests are disabled to speed up the build.

```bash
git checkout v3.11.3
mkdir build && cd build
cmake .. -DJSON_BuildTests=OFF
cmake --install .
# For non-root users:
# sudo cmake --install .
```

--------------------------------

### CANN Toolkit Environment Variables

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Details the environment variables that need to be set for using the CANN toolkit, including DDK_PATH and NPU_HOST_LIB, which point to the toolkit's installation directory and libraries.

```APIDOC
Environment Variables:

- DDK_PATH: Specifies the root directory of the Ascend CANN toolkit installation.
  Example: export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest

- NPU_HOST_LIB: Specifies the path to the host library directory within the CANN toolkit.
  Example: export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/lib64

- LD_LIBRARY_PATH: Should be updated to include the operator API library path after installing custom operators.
  Example: export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/opp/vendors/customize/op_api/lib:${LD_LIBRARY_PATH}
```

--------------------------------

### Install Files Configuration

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/sub_frameworklaunch/CMakeLists.txt

Defines files to be installed as part of the build process. This example installs 'sub_frameworklaunch.cpp' to a specific destination directory.

```cmake
install(FILES op_kernel/sub_frameworklaunch.cpp
        DESTINATION ${ASCEND_IMPL_OUT_DIR}/dynamic)
```

--------------------------------

### Installing Custom Operator Run Package

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Installs the compiled custom operator package (a .run file). It can be installed quietly to the default CANN package directory or to a specified custom path using the --install-path argument.

```bash
./build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run --quiet
./build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run --quiet --install-path=${install_path}
```

--------------------------------

### Environment Configuration for Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/pooling/adaptive_max_pool3d_grad/examples/AclNNInvocationNaive/README.md

Sets necessary environment variables for running the example, specifically DDK_PATH and NPU_HOST_LIB, which point to the Ascend toolkit installation.

```bash
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib
```

--------------------------------

### Setting Environment Variables for CANN Toolkit

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Sets up the necessary environment variables for the Ascend CANN toolkit, depending on the installation path (root, non-root, or custom). Also includes exporting DDK_PATH and NPU_HOST_LIB for custom paths.

```bash
source /usr/local/Ascend/ascend-toolkit/set_env.sh
```

```bash
source $HOME/Ascend/ascend-toolkit/set_env.sh
```

```bash
source ${install_path}/ascend-toolkit/set_env.sh
```

```bash
export DDK_PATH=${install_path}/ascend-toolkit/latest
```

```bash
export NPU_HOST_LIB=${install_path}/ascend-toolkit/latest/lib64
```

--------------------------------

### Configuring CMake for Custom Operator Compilation

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Shows example configurations within the CMakePresets.json file for setting the ASCEND_COMPUTE_UNIT and ASCEND_CANN_PACKAGE_PATH variables, which are crucial for custom operator compilation.

```json
{
    "ASCEND_COMPUTE_UNIT": {
        "type": "STRING",
        "value": "ascendxxxy"
    },
    "ASCEND_CANN_PACKAGE_PATH": {
        "type": "PATH",
        "value": "/usr/local/Ascend/ascend-toolkit/latest"
    }
}
```

--------------------------------

### Compile and Run AclNN Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conv/conv2d_backprop_input_v2/examples/AclNNInvocationNaive/README.md

Steps to compile and run the AclNN single-operator invocation example. This involves navigating to the example directory, setting environment variables, and then building and executing the test.

```bash
cd ${git_clone_path}/cann-ops/src/conv/conv2d_backprop_input_v2/examples/AclNNInvocationNaive

# Environment variable configuration (example for arm)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

# Build and run
mkdir -p build
cd build
cmake .. && make
./execute_test_op
```

--------------------------------

### Cloning cann-ops Repository

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Clones the cann-ops project source code from Gitee using the git command-line tool.

```bash
git clone https://gitee.com/ascend/cann-ops.git
```

--------------------------------

### ST Test Configuration File Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/xlogy/tests/st/README.md

Example configuration file (msopst.ini) used for ST testing, specifying parameters like atc_singleop_advance_option and HOST_ARCH.

```ini
# Example msopst.ini content
# atc_singleop_advance_option = ...
# HOST_ARCH = ...

```

--------------------------------

### Run AclNN Example Operator

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conversion/pad_v3_grad_replication/examples/AclNNInvocationNaive/README.md

Instructions to compile and run the AclNN operator example. This involves creating a build directory, configuring with CMake, compiling with make, and executing the generated binary.

```bash
mkdir -p build
cd build
cmake .. && make
./execute_test_op
```

--------------------------------

### Example Execution: Build and Run AngleV2 Operator

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/angle_v2/examples/AclNNInvocationNaive/README.md

Steps to compile and run the AngleV2 operator example application. This involves navigating to the example directory, setting environment variables, generating test data, compiling the C++ code using CMake and Make, and executing the compiled application.

```bash
cd ${git_clone_path}/cann-ops/src/math/angle_v2/examples/AclNNInvocationNaive

# Environment variable configuration (example for arm)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/lib64

# Example execution
python gen_data.py
mkdir -p build
cd build
cmake .. && make
./execute_angle_v2_op
cd ..
```

--------------------------------

### Run Operator Example with run.sh

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/angle_v2/examples/AclNNInvocationNaive/README.md

A convenience script to automate the compilation and execution process for the AngleV2 operator example. It encapsulates the build and run commands for a streamlined workflow.

```bash
bash run.sh
```

--------------------------------

### msopst.ini Configuration Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/fill/tests/st/README.md

An example configuration file for msOpST, specifying parameters like `atc_singleop_advance_option` and `HOST_ARCH`. These settings are used to tailor the test execution to the specific hardware environment.

```ini
; Example configuration for msopst.ini
; Modify atc_singleop_advance_option and HOST_ARCH based on your execution machine's architecture.

[common]
; ... other common configurations ...

[msopst]
atc_singleop_advance_option = ""
HOST_ARCH = ""
; ... other msopst specific configurations ...
```

--------------------------------

### Compile and Deploy Operator Package

Source: https://github.com/seelevolle/cann-ops/blob/master/src/quant/dynamic_quant/README.md

Provides shell commands to navigate to the repository, build the custom operator package, and deploy the generated installer.

```bash
cd ${git_clone_path}/cann-ops
```

```bash
bash build.sh
```

```bash
bash build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run
```

--------------------------------

### Execute Example via Script

Source: https://github.com/seelevolle/cann-ops/blob/master/src/index/feeds_repeat/examples/AclNNInvocationNaive/README.md

Run the example application using the provided run.sh script, which automates the build and execution process.

```bash
bash run.sh
```

--------------------------------

### AclNNInvocationNaive Example Execution

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conversion/pad_v4_grad/examples/AclNNInvocationNaive/README.md

Steps to compile and run the AclNNInvocationNaive example. This includes navigating to the example directory, setting necessary environment variables, and executing the build and run commands.

```bash
cd ${git_clone_path}/cann-ops/src/math/pad_v4_grad/examples/AclNNInvocationNaive

# Environment variable configuration (example for arm)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

# Build and run
mkdir -p build
cd build
cmake .. && make
./execute_add_op
```

--------------------------------

### Example Execution Commands

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/fast_gelu_grad/examples/AclNNInvocationNaive/README.md

Commands to navigate to the example directory, configure environment variables, and build/run the FastGeluGrad operator example.

```Bash
cd ${git_clone_path}/cann-ops/src/math/fast_gelu_grad/examples/AclNNInvocationNaive

export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

mkdir -p build
cd build
cmake .. && make
./execute_fast_gelu_grad_op
```

--------------------------------

### Build and Deploy Cross Operator

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/cross/README.md

Commands to compile the CANN custom operator and deploy the operator package. Requires the CANN software installation guide for environment setup.

```bash
cd ${git_clone_path}/cann-ops
bash build.sh
bash build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run
```

--------------------------------

### ST Test Execution Environment Setup

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/clip_by_value_v2/tests/st/README.md

Configures essential environment variables required for running System Tests (ST) with the msOpST tool. These variables point to the installation directory and host library.

```bash
export DDK_PATH=${INSTALL_DIR}
export NPU_HOST_LIB=${INSTALL_DIR}/{arch-os}/devlib
```

--------------------------------

### build.sh Script Parameters

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Lists the available parameters for the build.sh script, used for compiling custom operator packages. Includes options for specifying operator names and enabling unit tests.

```APIDOC
build.sh Script Usage:

- build.sh
  Description: Compiles all custom operator packages.

- build.sh -n <operator_name>
  Description: Compiles and packages a specific custom operator by its name.
  Example: bash build.sh -n add_custom

- build.sh -t
  Description: Compiles and executes unit tests for the operators.

- build.sh --help
  Description: Displays all available parameters and usage information for the build script.
```

--------------------------------

### Compile and Run Operator Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/select_reduce_max_d_sub_exp_reduce_sum_d_real_div/examples/AclNNInvocationNaive/README.md

Steps to compile and run the operator example application. This involves navigating to the example directory, setting environment variables, and then building and executing the application.

```bash
cd ${git_clone_path}/cann-ops/src/math/select_reduce_max_d_sub_exp_reduce_sum_d_real_div/examples/AclNNInvocationNaive

# Environment variable configuration (example for arm)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

# Example execution
mkdir -p build
cd build
cmake .. && make
./execute_select_reduce_max_d_sub_exp_reduce_sum_d_real_div_op
```

--------------------------------

### ST Test Execution Environment Setup

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/triu/tests/st/README.md

Configures essential environment variables required for running System Tests (ST) with the msOpST tool. These variables point to the installation directory and NPU host libraries.

```bash
export DDK_PATH=${INSTALL_DIR}
export NPU_HOST_LIB=${INSTALL_DIR}/{arch-os}/devlib
```

--------------------------------

### Build and Run Example Application

Source: https://github.com/seelevolle/cann-ops/blob/master/src/index/feeds_repeat/examples/AclNNInvocationNaive/README.md

Compile the example application using CMake and Make, then execute the compiled binary. This process automatically generates test data, compiles, and runs the AclNN example.

```bash
mkdir -p build
cd build
cmake .. && make
./execute_add_op
```

--------------------------------

### Run ForeachAtan Operator Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/foreach/foreach_atan/examples/AclNNInvocationNaive/README.md

Executes the ForeachAtan operator example. This script automatically generates test data, compiles and runs the aclnn example, and prints the results.

```Bash
cd ${git_clone_path}/cann-ops/src/foreach/foreach_atan/examples/AclNNInvocationNaive
bash run.sh
```

--------------------------------

### Compile and Run Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conversion/strided_slice_assign_v2/examples/AclNNInvocationNaive/README.md

Instructions for compiling and running the AclNNInvocationNaive example, including setting environment variables and executing the build and run commands.

```bash
# Navigate to the example directory
cd ${git_clone_path}/cann-ops/src/math/add_custom/examples/AclNNInvocationNaive

# Set environment variables (example for ARM)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

# Build and run the example
mkdir -p build
cd build
cmake .. && make
./execute_add_op

# Alternatively, use the provided run.sh script
# bash run.sh
```

--------------------------------

### Execute UpsampleTrilinear3dBackward Example via run.sh

Source: https://github.com/seelevolle/cann-ops/blob/master/src/image/upsample_trilinear3d_backward/examples/AclNNInvocationNaive/README.md

This script executes the `AclNNInvocationNaive` example for the UpsampleTrilinear3dBackward operator. It guides users to navigate to the example directory and then run the `run.sh` script, which automatically generates test data, compiles, runs the aclnn example, and prints the results.

```bash
cd ${git_clone_path}/cann-ops/src/image/upsample_trilinear3d_backward/examples/AclNNInvocationNaive
bash run.sh
```

--------------------------------

### CMakePresets.json Configuration Options

Source: https://github.com/seelevolle/cann-ops/blob/master/QuickStart.md

Describes key configuration options within the CMakePresets.json file used for building custom operators. These include specifying the target compute unit and the path to the CANN package.

```APIDOC
CMakePresets.json Configuration:

- ASCEND_COMPUTE_UNIT:
  Type: STRING
  Description: Specifies the Ascend compute unit (chip type) for compilation. This value can be queried using the 'npu-smi info' command (Name column).
  Example Value: "ascendxxxy"

- ASCEND_CANN_PACKAGE_PATH:
  Type: PATH
  Description: Specifies the installation path of the Ascend CANN development toolkit. This value can be queried using 'echo ${ASCEND_HOME_PATH}'.
  Example Value: "/usr/local/Ascend/ascend-toolkit/latest"
```

--------------------------------

### Deploy Operator Package

Source: https://github.com/seelevolle/cann-ops/blob/master/src/image/upsample_nearest_exact3d_grad/README.md

Command to install the compiled custom operator package onto the system. This typically involves running a generated installer script.

```bash
bash build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run
```

--------------------------------

### Build and Execute Example Operator

Source: https://github.com/seelevolle/cann-ops/blob/master/src/quant/dynamic_quant/examples/AclNNInvocationNaive/README.md

Commands to compile and run the AclNNInvocationNaive example. This involves creating a build directory, configuring the build with CMake, compiling the project using make, and then executing the compiled application.

```bash
mkdir -p build
cd build
cmake .. && make
./execute_test_op
```

--------------------------------

### Configure Environment Variables for Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/index/feeds_repeat/examples/AclNNInvocationNaive/README.md

Set necessary environment variables for the example execution, specifically DDK_PATH and NPU_HOST_LIB, which are required for Ascend toolkit components.

```bash
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib
```

--------------------------------

### Compiling and Running the Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conv/conv3d_backprop_input_v2/examples/AclNNInvocationNaive/README.md

Steps to compile and execute the Conv3DBackpropInputV2 operator example. This includes navigating to the example directory, setting necessary environment variables (DDK_PATH, NPU_HOST_LIB), building the project using CMake and Make, and running the compiled executable.

```bash
cd ${git_clone_path}/cann-ops/src/conv/conv3d_backprop_input_v2/examples/AclNNInvocationNaive

export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/lib64

mkdir -p build
cd build
cmake .. && make
./execute_conv3d_backprop_input_v2_op
cd ..

bash run.sh
```

--------------------------------

### Compile and Run Operator Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/eye_fp64/examples/AclNNInvocationNaive/README.md

Shell script to compile and run the EyeFp64 operator example. It handles environment setup and execution for testing the operator's functionality.

```bash
run.sh
```

--------------------------------

### Compile and Run AclNN Naive Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conversion/pad_v3_grad_replicate/examples/AclNNInvocationNaive/README.md

Steps to compile and execute the naive AclNN invocation example. This includes navigating to the example directory, setting necessary environment variables for the Ascend toolkit, building the project using CMake and Make, and running the compiled executable.

```bash
cd ${git_clone_path}/cann-ops/src/math/reflection_pad_3d_grad/examples/AclNNInvocationNaive
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib
mkdir -p build
cd build
cmake .. && make
./execute_test_op
```

--------------------------------

### Install PyTorch and Torch-NPU

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/add_custom/examples/CppExtensions/README.md

Installs specific versions of PyTorch and the torch-npu package, which are prerequisites for running the provided examples. These commands ensure compatibility with the CANN framework.

```bash
pip3 install torch==2.1.0
pip3 install torch-npu==2.1.0.post10
```

--------------------------------

### Example Execution Steps

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/inplace_fused_matmul_softmax_grad/examples/AclNNInvocationNaive/README.md

Instructions for compiling and running the aclnn example. This includes navigating to the example directory, setting environment variables, and executing the build and run commands.

```bash
# Navigate to example directory
cd ${git_clone_path}/cann-ops/src/math/inplace_fused_matmul_softmax_grad/examples/AclNNInvocationNaive

# Environment variable configuration (example for ARM)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

# Build and run
mkdir -p build
cd build
cmake .. && make
./execute_inplace_fused_matmul_softmax_op

# Alternatively, use the provided script
# bash run.sh
```

--------------------------------

### Example Test Case Definition File

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/rsqrt/tests/st/README.md

An example JSON file structure for defining test cases, including various data types and scenarios for operator testing.

```json
{
  "test_cases": [
    {
      "input": {
        "data": [1.0, 4.0, 9.0, 16.0]
      },
      "expected_output": {
        "data": [1.0, 0.5, 0.3333333333333333, 0.25]
      },
      "type": "float32"
    }
  ]
}
```

--------------------------------

### Compile and Run KlDivTargetBackward Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/kl_div_target_backward/examples/AclNNInvocationNaive/README.md

Shell script to compile and run the aclnn example for the KlDivTargetBackward operator. It includes steps for setting up environment variables and executing the compiled binary.

```bash
cd ${git_clone_path}/cann-ops/src/math/kl_div_target_backward/examples/AclNNInvocationNaive

export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/lib64

mkdir -p build
cd build
cmake .. && make
./execute_add_op
```

--------------------------------

### Build and Run Example Operator

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/is_finite/examples/AclNNInvocationNaive/README.md

Provides instructions to compile and run the IsFinite operator example. It involves navigating to the example directory, creating a build folder, configuring with CMake, compiling, and executing the generated binary.

```bash
cd ${git_clone_path}/cann-ops/src/math/is_finite/examples/AclNNInvocationNaive

mkdir -p build
cd build
cmake .. && make
./execute_is_finite_op
```

--------------------------------

### Build and Run Example Script

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/bev_pool/examples/AclNNInvocationNaive/README.md

Provides instructions and commands to compile and run the BevPool operator example. It includes setting environment variables, building the project using CMake and Make, and executing the compiled application.

```bash
cd ${git_clone_path}/cann-ops/src/math/bev_pool/examples/AclNNInvocationNaive

export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

mkdir -p build
cd build
cmake .. && make
./execute_add_op
```

--------------------------------

### Build and Install PyTorch NPU Plugin

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/add_custom/examples/PytorchInvocation/README.md

Copies custom kernel files, sets environment variables, builds the PyTorch NPU plugin, and installs it via pip.

```bash
cp -rf op_plugin_patch/*.cpp ${PYTORCH_DIR}/third_party/op-plugin/op_plugin/ops/v2r1/opapi
export DISABLE_INSTALL_TORCHAIR=FALSE
cd ${PYTORCH_DIR}
(bash ci/build.sh --python=${PYTHON_VERSION} --disable_rpc ; pip uninstall torch-npu -y ; pip3 install dist/*.whl)
```

--------------------------------

### Run UpsampleBicubic2dGrad Example Script

Source: https://github.com/seelevolle/cann-ops/blob/master/src/image/upsample_bicubic2d_grad/examples/AclNNInvocationNaive/README.md

This bash script guides users through navigating to the example code directory and executing the `run.sh` script. The `run.sh` script handles the compilation and execution of the aclnn example for the UpsampleBicubic2dGrad operator, automatically generating test data and printing the results.

```bash
cd ${git_clone_path}/cann-ops/src/image/upsample_bicubic2d_grad/examples/AclNNInvocationNaive
bash run.sh
```

--------------------------------

### Install ATK Dependencies

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/cast/tests/atk/README.md

Installs the necessary ATK wheel package, which is a prerequisite for running the tests. Ensure all required dependencies are met before execution.

```bash
pip3 install ATK*.whl
```

--------------------------------

### CMakeLists.txt

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/kl_div_target_backward/examples/AclNNInvocationNaive/README.md

CMake build script for compiling the KlDivTargetBackward operator example.

```cmake
CMakeLists.txt
```

--------------------------------

### Run ForeachMaximumScalar Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/foreach/foreach_maximum_scalar/examples/AclNNInvocationNaive/README.md

Instructions to compile and run the ForeachMaximumScalar example. This involves navigating to the example directory and executing a shell script that handles data generation, compilation, and execution.

```bash
cd ${git_clone_path}/cann-ops/src/foreach/foreach_maximum_scalar/examples/AclNNInvocationNaive
bash run.sh
```

--------------------------------

### ST Test Case Definition Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/xlogy/tests/st/README.md

Example JSON file (Xlogy_case_alltype.json) defining test cases for operator testing, including input data, expected output, and boundary conditions.

```json
{
  "test_cases": [
    {
      "input": {
        "x": [1.0, 2.0, 3.0],
        "y": [2.0, 3.0, 4.0]
      },
      "expected_output": {
        "z": [2.0, 6.0, 12.0]
      },
      "description": "Basic xlogy test"
    }
    // ... other test cases
  ]
}
```

--------------------------------

### Example Execution Commands

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/histogram_v2/examples/AclNNInvocationNaive/README.md

Steps to compile and run the aclnn example for the HistogramV2 operator. This includes navigating to the example directory, setting necessary environment variables (DDK_PATH, NPU_HOST_LIB), building the project using CMake and Make, and executing the compiled application.

```bash
cd ${git_clone_path}/cann-ops/src/math/histogram_v2/examples/AclNNInvocationNaive

export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

mkdir -p build
cd build
cmake .. && make
./execute_test_op
```

--------------------------------

### AclNNInvocationNaive Directory Structure

Source: https://github.com/seelevolle/cann-ops/blob/master/src/quant/dynamic_quant/examples/AclNNInvocationNaive/README.md

Overview of the directory structure for the AclNNInvocationNaive example, highlighting key files such as CMakeLists.txt for compilation, gen_data.py for data generation, main.cpp for the application entry point, run.sh for script execution, and verify_result.py for result comparison.

```text
├── AclNNInvocationNaive
│   ├── CMakeLists.txt      // 编译规则文件
│   ├── gen_data.py         // 算子期望数据生成脚本
│   ├── main.cpp            // 单算子调用应用的入口
│   ├── run.sh              // 编译运行算子的脚本
│   └── verify_result.py    // 计算结果精度比对脚本
```

--------------------------------

### ST Test Configuration Example (msopst.ini)

Source: https://github.com/seelevolle/cann-ops/blob/master/src/index/scatter_add_with_sorted/tests/st/README.md

An example INI configuration file for msOpST, used to specify parameters like atc_singleop_advance_option and HOST_ARCH based on the execution machine's architecture.

```ini
; Example msopst.ini content
; atc_singleop_advance_option = "--enable_small_batch=1 --disable_fusion=1"
; HOST_ARCH = "aarch64"
```

--------------------------------

### Navigate to Example Code Directory

Source: https://github.com/seelevolle/cann-ops/blob/master/src/index/feeds_repeat/examples/AclNNInvocationNaive/README.md

Change the current directory to the location of the example code for the AclNNInvocationNaive application.

```bash
cd ${git_clone_path}/cann-ops/src/math/add_custom/examples/AclNNInvocationNaive
```

--------------------------------

### Compile PyTorch Plugin and Install

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/kl_div_target_backward/examples/PytorchInvocation/README.md

Compiles the PyTorch NPU plugin, installs it, and prepares the environment for running custom operators. This involves copying kernel files and building the wheel package.

```bash
cp -rf op_plugin_patch/*.cpp ${PYTORCH_DIR}/third_party/op-plugin/op_plugin/ops/v2r1/opapi
export DISABLE_INSTALL_TORCHAIR=FALSE
cd ${PYTORCH_DIR}
(bash ci/build.sh --python=${PYTHON_VERSION} --disable_rpc ; pip uninstall torch-npu -y ; pip3 install dist/*.whl)
```

--------------------------------

### ST Test Case Definition Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/norm/add_rms_norm_cast/tests/st/README.md

An example JSON file structure for defining ST test cases, including input data, expected outputs, and boundary conditions for operator testing.

```json
{
  "AddRmsNormCast_case.json": "// Test case definition file example (generated by 8.0.RC3.alpha003 version)"
}
```

--------------------------------

### Compile and Run Example (Bash)

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conversion/reflection_pad3d_grad/examples/AclNNInvocationNaive/README.md

Shell script commands to compile and execute the AclNNInvocationNaive example for the ReflectionPad3dGrad operator. This includes setting environment variables, building the project with CMake, and running the compiled executable.

```bash
# Navigate to the example directory
cd ${git_clone_path}/cann-ops/src/math/reflection_pad_3d_grad/examples/AclNNInvocationNaive

# Set environment variables (example for arm)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

# Build the project
mkdir -p build
cd build
cmake .. && make

# Execute the compiled test
./execute_test_op
```

--------------------------------

### Compile and Run Operator Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/pre_layer_norm/examples/AclNNInvocationNaive/README.md

Shell script and commands to compile the example application and run the operator test cases.

```Bash
cd ${git_clone_path}/cann-ops/src/math/pre_layer_norm/examples/AclNNInvocationNaive

mkdir -p build
cd build
cmake .. && make
./execute_add_op {case-id}
```

```Bash
bash run.sh {case-id}
```

--------------------------------

### Test Case Definition Example (JSON)

Source: https://github.com/seelevolle/cann-ops/blob/master/src/index/scatter_add_with_sorted/tests/st/README.md

An example JSON file defining test cases for the scatter_add_with_sorted operator, typically used by the msOpST tool to specify inputs, outputs, and test scenarios.

```json
{
  "test_cases": [
    {
      "input_data": [
        {
          "name": "input_tensor",
          "data_type": "float32",
          "shape": [1, 3, 224, 224]
        },
        {
          "name": "indices",
          "data_type": "int32",
          "shape": [100],
          "value": "range(100)"
        },
        {
          "name": "updates",
          "data_type": "float32",
          "shape": [100],
          "value": "random"
        }
      ],
      "expected_output": {
        "name": "output_tensor",
        "data_type": "float32",
        "shape": [1, 3, 224, 224]
      }
    }
  ]
}
```

--------------------------------

### Build and Deploy Operator Package

Source: https://github.com/seelevolle/cann-ops/blob/master/src/foreach/foreach_mul_list/README.md

Provides the necessary bash commands to compile the custom operator package and deploy it to the system. This includes navigating to the project directory, executing the build script, and running the deployment installer.

```bash
cd ${git_clone_path}/cann-ops
bash build.sh -n foreach_mul_list
bash build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run
```

--------------------------------

### Get Soc Version

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conversion/pad_v4_grad/tests/st/README.md

Retrieves the Soc Version by querying the NPU device information.

```bash
npu-smi info
```

--------------------------------

### Operator Package Compilation and Deployment

Source: https://github.com/seelevolle/cann-ops/blob/master/src/foreach/foreach_sigmoid/README.md

Instructions for compiling and deploying the custom operator package. This involves navigating to the repository, executing a build script, and running the deployment package.

```bash
cd ${git_clone_path}/cann-ops
bash build.sh -n foreach_sigmoid
bash build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run
```

--------------------------------

### ST Test Case Definition Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/norm/inplace_add_rms_norm/tests/st/README.md

An example JSON file structure for defining ST test cases. This file typically includes input data, expected outputs, and configuration for operator tests.

```json
{
  "test_cases": [
    {
      "case_name": "positive_case_1",
      "input": {
        "data": "[[1.0, 2.0], [3.0, 4.0]]",
        "axis": 0
      },
      "expected_output": {
        "output_data": "[[1.0, 2.0], [3.0, 4.0]]"
      },
      "params": {
        "dtype": "float32"
      }
    }
  ]
}
```

--------------------------------

### Build and Deploy Isamin Operator

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/isamin/README.md

Provides the necessary bash commands to compile the custom Isamin operator and deploy the generated operator package onto the target system. This includes navigating to the repository and executing build scripts.

```bash
cd ${git_clone_path}/cann-ops
bash build.sh -n isamin
bash build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run
```

--------------------------------

### msopst.ini Configuration Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/matmul/gemm_v2/tests/st/README.md

Example configuration file for msOpST, used to define test parameters and environment settings. Key parameters like atc_singleop_advance_option and HOST_ARCH may need adjustment based on the execution machine.

```ini
; st测试配置文件
; ... other configurations ...
atc_singleop_advance_option = 
HOST_ARCH = 
```

--------------------------------

### Test Case File Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/triu/tests/st/README.md

An example of a test case definition file (JSON format) used for System Tests (ST). This file specifies inputs, expected outputs, and boundary conditions for operator testing.

```json
{
  "test_cases": [
    {
      "case_name": "triu_basic_test",
      "input": {
        "x": [[1, 2, 3], [4, 5, 6], [7, 8, 9]],
        "k": 0
      },
      "expected_output": {
        "y": [[1, 0, 0], [4, 5, 0], [7, 8, 9]]
      }
    }
  ]
}
```

--------------------------------

### Compile and Deploy Operator Package

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/isamax/README.md

Commands to compile the custom operator package and deploy it. This involves navigating to the repository, executing the build script, and running the deployment package.

```bash
cd ${git_clone_path}/cann-ops
```

```bash
bash build.sh -n isamax
```

```bash
bash build_out/CANN-custom_ops-<cann_version>-linux.<arch>.run
```

--------------------------------

### Compile and Run Matmul Operator Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/matmul_api_constant/examples/AclNNInvocationNaive/README.md

Instructions to compile and run the matmul_api_constant operator example. This involves setting up environment variables, creating a build directory, configuring with CMake, compiling with Make, and executing the compiled binary.

```bash
cd ${git_clone_path}/cann-ops/src/math/matmul_api_constant/examples/AclNNInvocationNaive

# Environment variable configuration (example for arm)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

# Compilation and execution
mkdir -p build
cd build
cmake .. && make
./execute_matmul_api_constant_op
```

```bash
bash run.sh
```

--------------------------------

### Run Operator Script

Source: https://github.com/seelevolle/cann-ops/blob/master/src/conv/conv2d_backprop_input_v2/examples/AclNNInvocationNaive/README.md

A shell script that automates the compilation and execution process for the operator example.

```bash
# run.sh
# Script for compiling and running the operator
```

--------------------------------

### Bash Script for Running GreaterEqual Operator Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/greater_equal/examples/AclNNInvocationNaive/README.md

Provides instructions and commands to compile and run the aclnnGreaterEqual operator example. It includes navigating to the example directory, setting environment variables for Ascend toolkit, building the project using CMake and Make, and executing the compiled binary.

```bash
cd ${git_clone_path}/cann-ops/src/math/greater_equal/examples/AclNNInvocationNaive

# 环境变量配置 (以arm为例)
export DDK_PATH=/usr/local/Ascend/ascend-toolkit/latest
export NPU_HOST_LIB=/usr/local/Ascend/ascend-toolkit/latest/aarch64-linux/devlib

# 样例执行
mkdir -p build
cd build
cmake .. && make
./execute_greater_equal_op

# 或者使用run.sh脚本
bash run.sh
```

--------------------------------

### ST Test Case Definition Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/norm/dua_quantize_add_layer_norm/tests/st/README.md

An example JSON file structure for defining System Test (ST) cases. This file typically includes various input data, expected outputs, and boundary conditions for operator testing.

```json
{
  "version": "1.0",
  "test_cases": [
    {
      "name": "basic_add_layer_norm",
      "description": "Test with typical input values.",
      "input": {
        "x": [1.0, 2.0, 3.0, 4.0],
        "gamma": 1.0,
        "beta": 0.0,
        "epsilon": 1e-5
      },
      "expected_output": {
        "y": [0.4472135954999579, 0.8944271909999159, 1.3416407864998738, 1.7888543819998317]
      },
      "type": "normal"
    }
  ]
}
```

--------------------------------

### Example Directory Structure

Source: https://github.com/seelevolle/cann-ops/blob/master/src/math/clip_by_value_v2/examples/AclNNInvocationNaive/README.md

Overview of the directory structure for the AclNNInvocationNaive example, highlighting key files like CMakeLists.txt, data generation scripts, main application entry point, run script, and verification script.

```text
├── AclNNInvocationNaive
│   ├── CMakeLists.txt      // 编译规则文件
│   ├── gen_data.py         // 算子期望数据生成脚本
│   ├── main.cpp            // 单算子调用应用的入口
│   ├── run.sh              // 编译运行算子的脚本
│   └── verify_result.py    // 计算结果精度比对脚本
```

--------------------------------

### GemmV2 Test Case Definition Example

Source: https://github.com/seelevolle/cann-ops/blob/master/src/matmul/gemm_v2/tests/st/README.md

Example JSON file defining test cases for the GemmV2 operator. This file typically includes various input data types, expected outputs, and boundary conditions for comprehensive testing.

```json
// GemmV2_case_all_type.json
// {
//   "test_cases": [
//     {
//       "input_data": {
//         "input_x": "...",
//         "input_y": "..."
//       },
//       "expected_output": "...",
//       "boundary_conditions": "..."
//     }
//   ]
// }
```