onnxruntime

eea3806d - model parallel refinement (#6244)

model parallel refinement (#6244) * Megatron Transformation as a seperate step * remove useless header * clang formating * Re-Structure megatron transformer for subsquent changes * fix comments