v1.5.0

Added

• New "Run in Separate Process" option that runs optimization and visualization in a dedicated worker process, keeping Python out of the Rhino process
  • Enabled by default
  • Supports both the normal optimization path and Human-in-the-Loop
  • Automatically falls back to the previous behavior when the worker is not available (e.g. on macOS)
• New Convert Result to Tunny Log component that converts optimization results from other plugins into a Tunny log file (.log)
• Support Wire display type: default, faint, hidden
• LAN Zoo license support via the TunnyZoo plugin
  • Supports machine-locked licenses and seat count
  • Distinguishes LAN Zoo leases from Cloud Zoo leases in the license status display
• GP (Optuna) sampler now supports selecting the acquisition function (UCB/EI/PI/Thompson Sampling)
• New Samplers
  • Restart CMA-ES Sampler
  • CMA-ES with Quasi-Random Refinement Sampling
• Tunny Dashboard
  • New standalone Tunny Dashboard launched via the TunnyRunTunnyDashboard Rhino command, the "Run Tunny Dashboard..." menu item, and the "Tunny Dashboard" ribbon button on the Visualize tab
  • Installed automatically on first use
  • Requires a valid license to run

Changed

• NSGA-III now auto-selects dividing_parameter from objective count and population size when set to AUTO/null
• Renamed the TunnyRunDashboard Rhino command to TunnyRunOptunaDashboard to distinguish it from the new Tunny Dashboard
• Renamed the Visualize tab's "Browser" ribbon group to "Dashboard"
• Improved UI responsiveness during optimization by making progress updates non-blocking
• Checking for a new version no longer hangs indefinitely if the network is slow or unresponsive
• The Python 3 component warning now suggests enabling "Run in Separate Process" to avoid the conflict, and is no longer shown when that option is already enabled

Removed

• Removed the unused Pruner feature and its related UI

Fixed

• Improved Python environment reinstall behavior on Windows, including notifying users when a Rhino restart is required
• When Python environment initialization fails, Tunny now shuts down cleanly and recommends reinstalling the Python environment from the application menu
• Once the Python interpreter has been shut down in a Rhino session it can't be restarted, so Tunny now asks the user to restart Rhino instead of leaving it half-initialized
• Installing the Python environment after Python had already been used in the same Rhino session used to leave the environment broken; Tunny now declines to install in that situation and installs automatically at the next start
• An install that failed or was interrupted partway is now retried at the next start instead of being treated as complete
• Failures during Python package installation are now detected and shown to the user, instead of reporting a broken environment as installed successfully
• If the Python runtime can't be initialized, Tunny now schedules a reinstall for the next start
• Plot generation errors no longer crash Rhino; an error message is shown instead
• Errors from the worker process (used by "Run in Separate Process") are now logged properly instead of being replaced by a secondary error
• Corrupted config files are now backed up before deletion and the user is notified, instead of being silently deleted
• Fixed CMA-ES and Restart CMA-ES samplers, which had stopped working due to upstream Optuna API changes
• Fixed CMA-ES and Restart CMA-ES using a hard-coded default population size instead of Optuna's automatic calculation
• Fixed garbled Japanese characters in the live chart
• Fixed a Rhino crash when restoring a Trial with a non-existent ID to Grasshopper
• Fixed some sampler settings being silently dropped when using "Run in Separate Process" (GA sampler mutation/crossover parameters; HEBO sampler's objective count)
• Fixed a crash on Rhino 7 (net48) when agreeing to the Terms of Use on first launch