Skip to content

Releases: PathmindAI/pathmind-webapp

Big update - NativeRL 1.2 support and more

22 Sep 20:15
61ad626
Compare
Choose a tag to compare

The version scheme is updated to match NativeRL's bump to 1.2

The major.minor version numbers will be kept in sync to show expected compatibility.

Other highlights of this release include:

  • Observation selection
  • New automatic reward variable names
  • User defined goals for projects
  • The production branch is renamed from master to prod

Support AnyLogic 8.6

27 Aug 15:55
Compare
Choose a tag to compare

Also includes the Simulation Metrics Uncertainty feature.

Hotfix for usage cap bug

04 Aug 22:01
dacd4e3
Compare
Choose a tag to compare

There was a bug that showed all users that they were hitting the usage limit cap for the day. This temporarily removes that check until we submit the actual fix.

#1936

Quick hotfix for reward variables

31 Jul 15:49
d8c8d11
Compare
Choose a tag to compare

Quick fix for reward variable names that would disappear. #1910

Tuples is in Prod!

29 Jul 22:19
a1dbd25
Compare
Choose a tag to compare

Tuples training, along with a month of other features is now released.

Highlights include simulation metrics.

Released July 29, 2020 3:15pm to Prod

Big release - Includes NativeRL 1.0.7

19 Jun 19:59
a369ccc
Compare
Choose a tag to compare

This is a month of updates. The UI will have been updated in many areas. NativeRL 1.0.7 is included and we update to using 36 cpu instances for training by default.

Trainings should be almost twice as fast.

May 20, 2020 - Increase allowed reward function size

20 May 23:27
4830f1e
Compare
Choose a tag to compare

This is a quick update to let a customer try their unusually large reward function.

It includes some UI formatting and bug fixes on the dashboard.

May 19, 2020 - Releasing the latest changes to production

19 May 21:27
292a328
Compare
Choose a tag to compare

The changes include a lot of tech debt refactoring. We also included changes that are required by Accenture and increase the stability of training by using NativeRL 1.0.6

May 12, 2020 - Early Stopping and more

12 May 21:12
ceb6b59
Compare
Choose a tag to compare
  • Upgrades NativeRL version to 1.0.5 that adjusts early stopping criteria
  • The Updater is split out to a separate service. This enables easier scalabilty.
  • Experiment page focuses more on the reward function
  • Show the model package name in the UI
  • Clean up the reward score chart by highlighting the highest score and removing the markers.
  • Fix bugs with resuming training
  • Paying down techdebt to increase performance and scalability.
  • More bug fixing
  • More automated tests
  • Lower infrastructure costs with NAT configuration

April 27, 2020

28 Apr 18:02
616075f
Compare
Choose a tag to compare
  • upgrades the NativeRL version being used to 1.03.
  • pbt will use 4 samples
  • UI won’t show reward score
  • Add model details to UI
  • Restyle the projects/project/model pages
  • bug fixes