#Feasability of implementing a mini inference engine

2 messages · Page 1 of 1 (latest)

tribal iris
#

Hello,I'm supposed to do a 2-personn project for my OOP course.It's supposed to be 3000+ lines of code .For a scale of the difficulty,the three proposed projects are a ray tracer,a tetris implementation(optionally a 3D version and online multiplayer) and a comic book reader.However we can also propose our own project.I'd like to make an AI-related one so i was thinking of making a small ONNX inference engine using c++.It would just implement a subset of operations like gemm,conv,maxpool,etc...
I was thinking i could use protobuf to parse an onnx model,interpret the graph,map onnx node strings to my implementations of some basic operations and then use CPU for the processing.
I was wondering if this is feasible for a project of this scale,and how hard it would be to implement optimizations that actually make use of some of c++'s low-level capabilities?

mellow deltaBOT
#

When your question is answered use !solved or the button below to mark the question as resolved.

Remember to ask specific questions, provide necessary details, and reduce your question to its simplest form. For tips on how to ask a good question use !howto ask.