Skip to content

Latest commit

 

History

History
77 lines (53 loc) · 3.64 KB

CHANGELOG.md

File metadata and controls

77 lines (53 loc) · 3.64 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

Features

  • OnnxExporter.add_metadata allows setting metadata to export with the policy. The key has to be a string, and the value has to be convertible to string.

Changes

New memory serialization format.

A new version of memory Table export has been introduced. This removes the need for cloudpickle, while focusing on restoring data. Old memories can still be imported, and you can also force the old format by passing version = TableSerializationFormat.Legacy in Table.store. The new format works by simply ingesting all data from the memory using the regular add_sequence, instead of filling the data stores directly. As part of this, the TableArray.store and TableArray.restore functions have new arguments to handle versioning.

  • New functions:
    • Column.configuration and Column.configure for save and load respectively
    • Strategy.clear to clear all state
    • Strategy.state and Strategy.load_state for save and load of strategy data.
    • Strategy.begin_simple_import and Strategy.end_simple_import to bookend the import process.
    • Strategy._in_simple_import to allow derived classes to bypass work while import is happening.

Other changes:

  • Now targetting torch version 1.12, up from 1.11.
  • OnnxExporter accepts a device argument to enable tracing on other devices.
  • FinalRewardTestCheck can now be configured with another key and to use windowed data.
  • begin_training has been split into restore_state followed by begin_training
  • CoverageBasedSampleStrategy has been added which allows memory sampling that prioritises unvisited experiences. This can speed up training.

Deprecations

  • emote.callbacks has been converted to a package. Future built-in callbacks will not be re-exported from emote.callbacks, and should instead be imported from their internal location.
  • emote.callbacks.LoggingMixin is now in the emote.mixins.logging module instead.

Bugfixes

  • Fix FeatureAgentProxy.input_names to use input_key when configured.
  • Callback.cycle can now be None
  • Fixed a deprecation warning with np.bool_ being used.

23.0.0 - 2023-03-03

Breaking

  • The minimum required Python version is now 3.9 (#87)
  • The torch-cpu feature has been renamed to torch as it wasn't limited to CPU-only variants. (#76)

Added

  • Enable exporting ONNX policies for Gaussian MLPs in emote.extra.onnx_exporter. This allows you to peridiocally write ONNX files to disk. (#80)
  • Add system performance logger in emote.extra.system_logger. This'll log memory and CPU usage to Tensorboard. (#81)
  • Add memory warmup waiter in emote.memory.memory to ensure the memory has enough data before starting to sample. This avoids the collector having to block the training loop when training async. (#78)

Changed

  • Our PDM plugin for torch management has been split off into a separate repository and published to PYPI. (#88)
  • Switch to PDM 2.3 as default version for testing (#62)
  • The input key used for Feature Agent Proxies can now be customized (#79)

22.0.0 - 2022-10-28

This is the initial release