v0.4.1

What's Changed

• Fixed bug preventing agent from receiving reward signals
• Simplification of publication process

Full Changelog: v0.4.0...v0.4.1

v0.4.0

What's Changed

Bug Fixes:

• Fix to Job object not subscriptable error
• Added error handling for new jobs
• Fixed training error message to say training and not testing
• Removed empty files
• Added fix to ensure model directory is created prior to saving model

Features

• list_policies(), list_algorithms(), list_encoders(), and list_configs() functions moved to be called from base nett
• Added new example scripts
• Changed the checkpoint save name format
• Simplified Monitor file saving

Brain:

• Added SimCLR as option for encoder
• Enabled reward argument
• Enabled embedding dimension to be customizable
• Enabled reward option (only supports extrinsic reward via "supervised")

Body:

• Enabled DVS option

Environment:

• New video recorder
• Simplified imports of configs

NETT:

• Addition of config argument to NETT.analyze()
• Added job_memory argument to NETT.run() to specify GPU memory reserved per job
• Removed broken CPU option for device argument for NETT.run()
• Added conditions argument to NETT.run() to support running a subset of all conditions
• Added error handling in NETT.analyze() for existence of necessary files
• Minor fix when average_in_episode_three_region return None
• Removed save of best model
• Made checkpoints optional with save_checkpoints arg (defaulting to False)
• Added option to adjust checkpoint saving frequency with checkpoint_freq argument in NETT.run()
• Changed default steps_per_episode from 200 to 1000 in NETT.run()
• Added synchronous option to NETT.run()

Full Changelog: v0.3.1...v0.4.0

v0.3.1

What's Changed

• Closes environments after force quits or other unexpected interruptions
• Documentation Cleanups

Full Changelog: v0.3...v0.3.1

v0.3

What's Changed

• Added CPU mode
• Allows for running on GPUs with insufficient memory
• Made loading bars run in parallel rather than on top of one another
• Fixed the total loading bar value to be accurate
• Fixed bug with /tmp/ files from other users blocking initialization of nett
• Allow for user-specified job memory reservation sizes
• Added "Getting Starting" section to online documentation
• Updated README to be more readable
• Updated research paper links to link to documentation website
• Updated online documentation formatting
• Added note in README regarding ability to watch the Unity environment in real time
• Switched docstrings over to Google-style for improved readability
• Created Job class for storing job information
• Added more thorough type hinting
• Restricted push to gh-pages job to main branch
• Created developer notes to include simple instructions for getting started on working on the codebase.

Full Changelog: v0.2...v0.3

v0.2

What's Changed

• Fixed installation issues
• Switched netts to nett
• Removed unnecessary dependencies
• Updated pyproject.toml
• Fixed hanging processes
• Updated README
• Create a documentation website
• Replaced ci.yml with docs.yml
• Added lint.yml
• Expanded .gitignore to include common R , macOS, and vs code ignore files
• Fix to License
• Added MANIFEST to include R and csv files in package
• Changed bug report templateFixed installation issues
• Switched netts to nett
• Removed unnecessary dependencies
• Updated pyproject.toml
• Fixed hanging processes
• Updated README
• Create a documentation website
• Replaced ci.yml with docs.yml
• Added lint.yml
• Expanded .gitignore to include common R , macOS, and vs code ignore files
• Fix to License
• Added MANIFEST to include R and csv files in package
• Changed bug report template

Full Changelog: https://github.com/buildingamind/NewbornEmbodiedTuringTest/commits/v0.2

v0.1.1

What's Changed

stable changes after 0.1 as of Feb 4 2023

Full Changelog: https://github.com/buildingamind/NewbornEmbodiedTuringTest/commits/v0.1.1

v0.1

Initial release

Full Changelog: https://github.com/buildingamind/NewbornEmbodiedTuringTest/commits/v0.1