General Pipeline for Kernel Tuner

[AndreSlavescu]

Project Description

• set up data pipeline to preprocess the output from GPT-4 turbo as a valid python program which defines the current state of the optimization strategy used to compile the resulting kernel in kernel tuner
• feed forward the input kernel parsed from the openAI API interface pipeline as input to kernel tuner and set up the process with the new optimization strategy
• Design a system to parse the timing results and memory usage results from the output of kernel tuner from both the prev state and the current kernel state and feed those back to the openAI interface to start the next iteration of optimizations. Feed forward the resulting .cu file to the openAI interface as well.