[TensorOperations]
precompile_workload = true