While we typically using HomogeneousWorkloadSpec to create homogeneous workloads (e.g., where min_input_tokens=max_input_tokens and min_output_tokens=max_output_tokens). It doesn't have to be used that way. It can actually be used to generate heterogeneous workloads just using simple uniform distributions.
Either we should:
Change the definition of HomogenousWorkloadSpec so that it is really homogeneous, or
Rename it to something like HomogeneousWorkloadSpec -> SyntheticWorkloadSpec and HeterogeneousWorkloadSpec -> RealisticWorkloadSpec
While we typically using
HomogeneousWorkloadSpec
to create homogeneous workloads (e.g., wheremin_input_tokens=max_input_tokens
andmin_output_tokens=max_output_tokens
). It doesn't have to be used that way. It can actually be used to generate heterogeneous workloads just using simple uniform distributions.Either we should:
HomogenousWorkloadSpec
so that it is really homogeneous, orHomogeneousWorkloadSpec
->SyntheticWorkloadSpec
andHeterogeneousWorkloadSpec
->RealisticWorkloadSpec