removing something unneeded from an example. first cut at hiopbbpy documentation

Author: thartland

doc/src/sections/hiopbbpy.tex

@@ -0,0 +1,26 @@
+%!TEX root = ../techrep_main.tex
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% HiOpBBpy
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{HiOpBBpy}
+
+\Hi offers a Bayesian optimization \Hibbpy solver to determine approximate global optimizers of black-box functions. Here, we detail how to install and utilize the package.
+
+\subsection{HiOpBBpy installation}
+
+\Hibbpy depends on the surrogate modeling toolkit (smt) and cyipopt which is a python wrapper of ipopt. In order to expose ipopt to cyipopt one can export the \verb~IPOPT_PATH~ and 
+
+\begin{verbatim}
+> export PKG_CONFIG_PATH=$PKG_CONFIG_PATH:$IPOPT_PATH/lib/pkgconfig
+> export PATH=$PATH:$IPOPT_PATH/bin
+\end{verbatim}
+
+To then install \Hibbpy from the home directory of \Hi:
+\begin{verbatim}
+> pip install .
+\end{verbatim}
+The ipopt libaries are then exposed:
+\begin{verbatim}
+> export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$IPOPT_PATH/lib
+\end{verbatim}

doc/src/techrep_main.tex

@@ -119,6 +119,7 @@
 \algnewcommand\Output{\item[\algorithmicoutput]}
 
 \newcommand{\Hi}{\texttt{HiOp}\xspace}
+\newcommand{\Hibbpy}{\texttt{HiOpBBpy}\xspace}
 \newcommand{\ignore}[1]{}
 
 \begin{document}
@@ -232,13 +233,14 @@
 
 \section{Introduction}
 
-This document describes the \Hi suite of HPC optimization solvers for some large-scale nonconvex nonlinear programming problems (NLPs). Four main classes of optimization problems are supported by \Hi. 
+This document describes the \Hi suite of HPC optimization solvers for some large-scale nonconvex nonlinear programming problems (NLPs). Five main classes of optimization problems are supported by \Hi. 
 \begin{itemize}
   \item \texttt{HiOp-Dense} supports NLPs with billions of variables with or without bounds but only limited number of
   constraints. This solver is a memory-distributed, MPI-based quasi-Newton interior-point solver using limited-memory approximations for the Hessians.
   \item \texttt{HiOp-Sparse} supports general sparse and large-scale NLPs sparse second-order derivatives. This functionality is similar to that of the state-of-the-art Ipopt~\cite{waecther_05_ipopt0}, but with additional features such as the inertia-free approach~\cite{Chiang2016}. The solver offers GPU acceleration via Nvidia CUDA Toolkit or AMD HIP Toolkit, and requires RAJA portability abstraction layer when GPU acceleration is enabled.
   \item \texttt{HiOp-MDS} supports NLPs that have dense and sparse blocks, for which a ``Newton'' interior-point solver is available together with a specialized, so-called mixed dense-sparse (MDS) linear algebra capable of achieving good performance on GPUs via Magma dense linear solver.
   \item \texttt{HiOp-PriDec} is an asynchronous memory-distributed optimization solver for two-stage stochastic programming problems. It implements a master-worker asynchronous scheduler based on MPI to improve load balancing. GPU acceleration can be achieved in solving each subproblem by \texttt{HiOp-MDS} or \texttt{HiOp-Sparse}.
+  \item \texttt{HiOpBBpy-BOAlgorithm} is a Bayesian optimization solver for low-dimensional expensive black-box functions subject to constraints. This solver makes use of batched Bayesian optimization methods and utilizes a master-worker scheduler in to evaluate the objective on the point batch. 
 \end{itemize}
 
 This document includes instructions on how to obtain and
@@ -970,11 +972,27 @@ \subsection{Compiling and linking your project with the \Hi library}
 In addition, a shared dynamic load library can be also built by using \texttt{HIOP\_BUILD\_SHARED} option with cmake. 
 
 
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% HiOpBBpy
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\input{sections/hiopbbpy.tex}
+
+ 
+
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 %% Solver options
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 \input{sections/solver_options.tex}
 
+
+
+
+
+
+
+
+
+
 %\input{old_stuff}
 
 \section{Licensing and copyright}

src/Drivers/hiopbbpy/BODriverEX.py

@@ -21,12 +21,6 @@
 from hiopbbpy.problems import BraninProblem, LpNormProblem
 from hiopbbpy.utils import MPIEvaluator
 
-# Get user input for the number of repetitions from command-line arguments
-if len(sys.argv) != 2 or int(sys.argv[1]) < 0:
-    num_repeat = 1
-else:
-    num_repeat = int(sys.argv[1])
-
 ### parameters
 n_samples = 5  # number of the initial samples to train GP
 theta = 1.e-2  # hyperparameter for GP kernel