Skip to content

Latest commit

 

History

History
75 lines (59 loc) · 2.62 KB

README.md

File metadata and controls

75 lines (59 loc) · 2.62 KB
Raw

ChemputerOptimizer

The repo contains the optimizer package for Chemputer platform to perform interactive chemical reaction optimizations. Checkout documentation for details about installation, usage and development. Below is a short guide.

Installation

  1. Clone the repo into desired folder.
git clone https://gitlab.com/croningroup/chemputer/chemputeroptimizer.git chemputeroptimizer

cd chemputeroptimizer
  1. Install Chemputer requirements if needed (XDL and AnalyticalLabware).
pip install -r requirements.txt
  1. Install ChemputerOptimizer.
pip install .

Usage

from optimizer import ChemputerOptimizer

co = ChemputerOptimizer('<xdl optimization procedure>', '<optimization graph>')
co.prepare_for_optimization(<'path to optimization config'>)
co.optimize('<chempiler instance>')

<xdl optimization procedure> is any valid .xdl procedure with some steps (targeted as optimization parameters) wrapped with OptimizeStep, e.g.

<OptimizeStep
  id="0"
  optimize_properties="{'mass': {'max_value': 0.24, 'min_value': 0.16}}">
  <Add
    reagent="copper(I)iodide"
    vessel="filter"
    mass="0.19 g"
    stir="False" />
</OptimizeStep>

Optionally some steps may be wrapped FinalAnalysis step with supported analytical method or interactive if you want to run optimization loop interactively.

<FinalAnalysis
  method="method">
  <HeatChill
    vessel="filter"
    temp="25 C"
    time="15 mins" />
</FinalAnalysis>

Optimizer Client

ChemputerOptimizer now supports interaction with Summit benchmarking framework (through SummitServer). List of available algorithms is stored in /client.py. To use them, run SummitServer main loop on chosen host and change the corresponding constants in client.py.

Features

v0.4.1

  • Catch up with xdl/chemputerxdl v1.5.

v0.4.2

  • Fixed incorrect OptimizeStep validation.

v0.4.3

  • Data is now saved using scientific notation (no data loss if target < 1e-04).
  • Multiple Analyze steps can now be used in procedure. Only first one will be update the optimization target.
  • Documentation update.

Development

Please check the issue list for the relevant things to do.

If you want to add another algorithm for the optimization, please follow the AbstractAlgorithm class documentation. Testing functions for algorithm evaluation can be found in /simulation.