Andes Neural Network Library offers a variety of functions, such as activation, convolution, fully connected layers, pooling, and other auxiliary operations.
- Andes Neural Network Library user manual
The user manual of Andes Neural Network Library can be found under docs/html. - Prebuilt library
The prebuilt library compiled with x86_64 GCC is available in theprebuilt/folder. - For any toolchains
- Steps to build the Andes Neural Network Library
Run build.sh with the compiler name as the argument. For example:
./build.sh "riscv32-unknown-elf-gcc"
Once the build is complete, a static library named libnn.a will be generated. For more usage details, please refer to the comments at the beginning of build.sh. - Steps to build the example
- Navigate to example folder
- Build the example with the following command (assumes riscv32-unknown-elf-gcc as the compiler):
riscv32-unknown-elf-gcc example.c -o example.out -lnn -L../ -I../Include
- Steps to build the Andes Neural Network Library
- For Andes toolchains
- Steps to build and install the Andes Neural Network Library multilib
Execute the build_default_library.sh script. For example:
./build_default_library.sh <RISCV32-UNKNOWN-ELF_TOOLCHAIN_PATH>/riscv32-elf/lib riscv32-elf <RISCV32-UNKNOWN-ELF_TOOLCHAIN_PATH>/riscv32-elf/include gccNote: Running build_default_library.sh will overwrite any existing header files and libraries from previous builds.
- Steps to build the example
- Navigate to example folder
- Build the example with the following command (assumes riscv32-unknown-elf-gcc as the compiler):
riscv32-unknown-elf-gcc example.c -o example.out -lnn
- Steps to build and install the Andes Neural Network Library multilib