Baremetal-NN

Baremetal-NN is a tool for converting PyTorch models into raw C codes that can be executed standalone in a baremetal runtime on research chips.

Note: After a discussion with @iansseijelly, we decided to switch to the simpler way of assuming array will be contiguous, and therefore directly use shape to index into elements, instead of the more generic strided access. The previous strided implementation can be access on the "strided" branch.

Run Test

Building for x86

first, we clean any previous builds

rm -rf ./build/

cmake . -S ./ -B ./build/ -D CMAKE_BUILD_TYPE=Debug
cmake --build ./build/ --target tests
./build/tests/tests

Building for RISC-V

first, we clean any previous builds

rm -rf ./build/

# make sure $RISCV is set
cmake . -D CMAKE_TOOLCHAIN_FILE=./riscv-gcc.cmake -S ./ -B ./build/ -D CMAKE_BUILD_TYPE=Debug
cmake --build ./build/ --target all
spike ./build/tests/tests.elf

Building for RISC-V with Vector Support

first, we clean any previous builds

rm -rf ./build/

# make sure $RISCV is set
cmake . -D CMAKE_TOOLCHAIN_FILE=./riscv-gcc.cmake -S ./ -B ./build/ -D CMAKE_BUILD_TYPE=Debug -D RVV=ON
cmake --build ./build/ --target all
spike --isa=rv64gcv_zicntr_zfh --varch=vlen:512,elen:32 ./build/tests/tests.elf

Running with FP16 support

cmake . -D CMAKE_TOOLCHAIN_FILE=./riscv-gcc.cmake -S ./ -B ./build/ -D CMAKE_BUILD_TYPE=Debug -D RVV=ON -D ZVFH=ON
cmake --build ./build/ --target all
spike --isa=rv64gcv_zicntr_zfh_zvfh --varch=vlen:512,elen:32 ./build/tests/tests.elf

Building for RISC-V with Gemmini

first, we clean any previous builds

rm -rf ./build/

cmake . -D CMAKE_TOOLCHAIN_FILE=./riscv-gcc.cmake -S ./ -B ./build/ -D CMAKE_BUILD_TYPE=Debug -D GEMMINI=ON
cmake --build ./build/ --target all
spike --extension=gemmini --misaligned ./build/tests/tests.elf

Building for K230 board

first, we clean any previous builds

rm -rf ./build/

cmake . -D CMAKE_TOOLCHAIN_FILE=./k230-gcc.cmake -S ./ -B ./build/ -G "Unix Makefiles" -D CMAKE_BUILD_TYPE=Debug -D RVV_ASM=ON
cmake --build ./build/ --target all

Cleaning build files

cmake --build ./build/ --target clean

Cleaning CMake files

rm -rf ./build/

Convert the model

python ./scripts/convert.py

the converter will dump out three files:

nn.h: stores the library definition.

operators.h: stores the operator definitions.

weights.h: stores the weights and biases of the network.

model.h: stores the code representation of the model forward pass.

Memory layout

Baremetal-NN uses the NHWC memory layout and supports up to 4-dimension tensor.

N: batch, H: height, W: width, C: channels

Code organization

The torch-like functions that operates on Tensor datatypes are under nn/functional.

The low-level implementations of kernels are under nn/impl/<device>.

For the low-level functions, the following naming convention is used:

void NN_operator_datatype(size_t n, <datatype *output_ptr, size_t increment>, <datatype *input_ptr, size_t increment>);

operator: the name of the operator, such as add, max.

dataType: the datatype of the operands, such as i8, u16, f32. If the datatype of the results and operands are different, it will be named <operand 0>_<operand 1>_..._to_<result 0>_...

output_ptr / input_ptr: the pointer to the data buffer with the correct type. The correct pointer type saves the repetitive casting within the function source code.

increment: the number of element to increment in order to access the next element in the buffer, in number of elements, not bytes. (e.g. for f32 type, increment of 1 will access next element starting from the next 4th byte, and hence the next contiguous fp32 number.)

Stats

Star History