Release Notes

Version 0.98.10

• Multi-node cluster setup added for Ray in RLMolecule examples

• Adding caching and multi-threading to a number of Chemistream GUI operations, which has improved UI response speed by up to >10x

• Workflows with SLURM scheduler now has button to cancel all queued and/or running jobs

• Compute session window has improved focus, where a user can move away from a cluster while building and do work in existing compute sessions

• Unique compute session label checking is now embedded

• Improved management of tensorboard ports in RLMolecule workflows

• Improved tab labeling for PDF files opened within compute session from JupyterLab

• Better checking for internet connectivity on startup

• Added unified access to selected molecular databases and a standardized internal Chemistream format for database info storage

• Updated compute session startup notebook to be clearer and more streamlined with storage interface notebook linked instead of open by default

• Checking for network connection on startup

• Improved behavior when re-connecting to a session after network connection is changed or interrupted

• Updated main dependencies: JupyterLab(4.0.10), ipywidgets(8.1.1)

• Updated MambaForge Python installation for compute images to version 23.10.0-0

• Fixed various minor bugs in the main Chemistream UI

• Updated the movie player for the SPPARKS off-lattice example

• Fixed behavior when reloading a new license

• Checking for correct Docker version when starting local compute session

• Adding more tooltips in the main Chemistream UI

• Adding checks when starting a new compute session that the selected AWS instance supports x86_64

• Unique cluster tag name is generated when starting a new compute session

Version 0.98.0

• Main Chemistream GUI improvements

  • Improved behavior when compute session setup fails, remote instances now terminate automatically thereby saving money

  • Further improvements in CI/CD pipelines for GUI and compute images

  • Better startup logic and error catching when loading license file and user settings

  • Improved UI design including improved error messaging and links to troubleshooting help page

  • Robust checking for internet connection before attempting remote compute session

  • Reorganizing compute session window and adding an ‘Instance type’ column

• Initial unit testing framework in place for main GUI

• Fixed bug for managing remote ssh keys on AWS

• Locking all dependencies for GUI and workflow images

• Improved re-connection behavior for remote compute sessions

• More robust compute session startup procedure, including purging old ssh key files

• Remote storage interface in workflows now connect to individual S3 accounts on AWS

• Updated demos/tutorials including new demos for NWChem and integrated charge analysis, ALFABET, reinforcement learning and BDE training data generation

• Improved error catching for storage interface in all workflows

• Single license file for application and compute images

• New off-lattice deposition model implemented in SPPARKS source

• Reinforcement learning (RL) workflow added. Includes user defined initial molecule, UI settings for hyperparameter scans and RLMolecule setup and molecule rendering for top rewards found

• Embedded tensorboard supported in RL workflows

• Improved documentation for RL workflow, descriptions of all menu settings added

• Improving handling of ports for tensorboard for local and remote compute sessions

• Simplifying dependencies in all compute images, reducing their size

• Updating SPPARKS version to include a recent bug fix branch

• Including license agreements for all underlying compute applications in and source code where required

Version 0.90.0

• Main GUI setup table now includes a dynamic cost estimator

• Improved clarity for all workflows implemented with a design of enabling disabled buttons/menus in all custom JupyterLab notebooks

• Simplified build dependencies for the images and main GUI

• updated to ipywidgets to 7.7.1, Jupyterlab 3.4.7 and Python 3.10

• SPPARKS image updated to version 19Sep22

• SPPARKS-ALD workflow has additional simulation parameters exposed including separate controls for:

  • species A,B pressure

  • species A,B deposition/desorption time

  • species A,B energy of surface reaction energies

  • number of adsorption/desorption cycles

• Makalii updates to improve setup stability

• Waihona storage interface to AWS S3/Azure Blob storage has improved clarity and stability with dynamic dropdown menus for up(down)load location in cloud compute sessions

• AI/ML images include setup for GPU instance (g3.4xlarge) at AWS

• Updated NREL AI/ML libraries NFP(0.3.12), GraphEnv(0.2.0) and RLMolecule(0.0.6) for the latest versions of TensorFlow(2.11.0) and Ray(2.10.0)

• Compute session docker images now include explicit memory allocation for /dev/shm in order to improve AI/ML library Ray performance

• Chemistream tools modules have been split up for reduced image sizes and faster startup

• Migrated from conda to mamba for faster development and testing

• Source code for main Chemistream GUI and compute images has been migrated from GitHub to GitLab

• User issue tracking, development and modern CI/CD build pipelines for the main GUI and compute images now implemented in private GitLab project

• New and improved product website (chemistream.io)

• Improved NWChem workflows for:

  • bond dissociation energy (BDE) calculations at scale ie accepting a CSV of SMILES strings

  • parsing of DDEC6 analysis files to seamlessly tabulate bond connectivity, bond-orders, partial charges and generate RDKit mol-object for additional analysis

  • using the NREL OPV molecule database to setup NWChem simulations for further analysis including DDEC6 bond/partial-charge analysis

• Bug fixes and improved robustness across all workflows

• Improved stability for embedded JSME drawing widget

• Prototype Chemistream embedded window for cloud resource management implemented

• NWChem setup menu now as a toggle for turning pre-opt HF/2-21G step on/off

• For local compute sessions, Chemistream performs a docker system clean on termination

• Simplified prototype for pricing models (still under development in cooperation with beta-testers) See Pricing Models for current details.

Version 0.80.0

• Improved ‘pre-optimization/embedding’ for molecules in DFT workflows (implemented with customized RDKit code).

• Improvements to Dalton & NWChem DFT workflows to accept CSV files with lists of SMILES strings.

• Improvements to Dalton & NWChem DFT workflows to improve the geometry optimization for organic molecules. There is now an explicit HF/3-21G calculation for both codes before proceeding to a user-requested functional and basis set to further refine geometry optimization.

• Adding parsing of polarizability and dipole moments for all DFT workflows.

• Enabled including solvent models in Dalton and NWChem DFT workflows.

• Adding menu choices to simplify control of level of accuracy during geometry optimization for NWChem workflows.

• Adding viz for HOMO/LUMO molecular orbitals for NWChem DFT workflows.

• Adding final molecular configuration output to mol2 files for DFT workflows.

• Integrating bonding and partial-charge analysis (using DDEC6) into NWChem DFT workflows.

• Toggle between default input template and user defined inputs for DFT workflows

• Formatting improvements for all workflows including - file manager window default width - instructions for starting terminals and better menu formatting

• Better startup instructions for user settings including a prototype UI for interacting with S3 storage at AWS. This permanent storage will be used for downloading workflows, checkpointing calculations and saving user-generated results.

• Reworking user account setup

  • AWS security group settings configured automatically

  • AWS ssh-key pairs managed automatically

  • default, reserved user settings account (‘techx’) can be used for connecting to common remote compute resources (prototype of collaborative platform mode)

  • new UI dialog for loading a user’s custom settings file

  • Automated AWS user-accounts through new Chemistream Tech-X website

• Improving AWS billing by using internally generated institution and user accounts at AWS by utilizing their organizational tools.

• Improved viz speed for ALD SPPARKS workflow.

• Improved model selection and setup for ALD SPPARKS workflow.

• Adding internet download/upload speed check in launchpad startup window.

• Improved stability for Mac/Windows installers (including working with conda constructor developers).

• Improved behavior for re-connecting to remote compute sessions when internet connections are changed or temporarily drop-out.

• Adding embedded Tensorboard capability for AI/ML workflows.

• Reworking AI/ML image dependencies to include the ‘Ray’ project. This is part of NREL’s refactoring of RL capability with the GraphEnv project. A new image is now available through Chemistream utilizing the GraphEnv NREL library.

Version 0.60.0

• Updating LAMMPS image to version 29Oct2020 and the corresponding LAMMPS-AWESEMMD

• New prototypes for NWChem workflow menus in NWChem NEXT examples

• Prototype of new launchpad PySide2 application for starting cloud resources in place.

• New quantum chemistry code DALTON image available

• Major revision to Chemistream starting application:

  • JupyterLab local manager session is deprecated

  • Startup GUI now manages license file installation and cloud startup functions

  • Multiple remote compute sessions can now be managed through startup GUI

  • Improved error handling throughout the startup process

• Reworking of compute session workflows for ALFABET, NWChem and Dalton to use ipywidgets for all functions

• Improving startup user instructions for all remote workflows

• New streamlined workflows for Dalton and NWChem for

  • scaling studies

  • parsing HOMO/LUMO values

  • series of runs given input SMILES strings

  • calculating hyperpolarizabilities

• Implemented multi-stage builds for Docker images resulting in a further 50% size reduction

• Updating panel and panel-chemistry dependencies resulting in faster app/image installs

• Updating to JupyterLab=3.3.0

• Using jupyterlab_scenes module to enable auto-execution for turnkey workflows

Version 0.20.0

• Reworked Chemistream application build system so pyinstaller no longer used, thereby resulting in smaller installers

• Streamlined startup page on JupyterLab: Manager tab

• Adding menu selection for new RLMolecule (from NREL) image

• Adding menu selection for new SPPARKS image with proprietary ALD model pre-compiled

• Adding menu selection for new NWChem image with the DDEC6 and Molden packages added

• Maintaining separate local ports for remote cloud compute session and local compute session

• Fixing temporary directory scratch space for NWChem sims

• Improved docker desktop setup for ‘local’ compute session across platforms

• Improved checks for installed docker and docker engine start for ‘local’ compute session

• Updating LAMMPS and SPPARKS examples for better results parsing

• Improved visualization for SPPARKS diffusion 3d example

• Reworking of dependencies in Chemistream installer. The installer depends on Makalii and Jupyterlab only. All compute modules and codes are moved to Docker images available from Tech-X repo. All installers are nearly half as large.

• Installer setup and configuration now ~2-3x faster due to simplified dependencies.

• All Tech-X Docker images have been reworked for fewer dependencies and are smaller

• JSME editor is now embedded in Jupyterlab notebooks, available through the Chemistream RDKitBase class

Version 0.12.1

• Updating to conda Python 3.8

• Updating to JupyterLab > 3

• Updating ipywidgets to latest version that simplifies and speeds up installation

• Updating pyinstaller and constructor packages to latest version (these can use conda python)

• Adding menu selection for new GROMACS image

• Using development version of NGLView that simplifies and speeds up installation

• Removing STREAMM dependency from Chemistream module as this is only used in remote compute sessions. This has decreased the size of installers on all platforms

• Adding ability to start a ‘local’ compute session so users can test workflows without need for configuring cloud accounts. This is valid on MacOS, Linux and Windows

• Improving stability of setup of compute session

• New example modules for BDE training model

• Bug fixes for storage (Waihona) interface methods

Version 0.9.12

• Removing py3Dmol dependencies and methods from ChemVis module

• Adding NGLView dependencies and migrating from py3Dmol

• New methods for NGLView added to ChemVis module

• Adding example directories for rDock and QMCPack

Version 0.9.10

• Upgrading to JupyterLab v 2.1.0

• Examples for remote storage (eg AWS S3 and Azure Blob)

• Adding RDKit dependency for a base class RDKitBase that is used by several examples’ derived classes.

• Adding example for downloading OPV database info and analyzing with RDKit

• Adding extra setup python files to tailor installs for app/cloud and (un)licensed versions

• Freezing Miniconda version to version Miniconda3-py37_4.8.3-platform-x86_64.sh”

Version 0.9.8

• Examples for machine learning, image recognition for ‘Digits’ and ‘Fashion’

• Example for neural net prediction of bond dissociation energies using Alfabet

• Example for complex P3HT workflow

Version 0.9.1

• Cloud management for Amazon and AWS

• Initial development of Jupyterlab framework using ipywidgets

• Examples for scaling with NWChem, LAMMPS and SPPARKS

• Cloud images all enabled with generation of multi-node HPC cluster

• Cloud images all enabled with creation of NFS shared directory