Explore projects

Parametric floating-point unit with support for standard RISC-V formats and operations as well as transprecision formats.

Aidge's quantization module, for both PTQ and QAT, written in C++

Aidge's ONNX import/export module, written in Python, required for ONNX interoperability

Aidge's TensorRT export module. It act as a thin wrapper around NVidia's tools with easy custom operator insertion

Aidge's reference C++ export module, required for generating standalone C++ static compute graph

Aidge's core module, written in C++, always required

Aidge's CUDA backend, written in C++/CUDA and using NVidia's CuDNN and CuBLAS libraries. Required for training and inference on NVidia's GPUs

Aidge's reference CPU backend, written in C++, not optimized for performances

Aidge meta-repository, which includes the main modules of the Aidge framework as Git submodules, plus the documentation and tutorials.