Skip to content

Latest commit

 

History

History
83 lines (72 loc) · 4.02 KB

CHANGELOG.md

File metadata and controls

83 lines (72 loc) · 4.02 KB
Raw

0.1.0

Initial release

0.1.1

Fix missing individual metrics for translation loss.

0.1.2

Fix wrong module used to compute the cycle losses. Don't do cycle with the same domain as target and source.

0.2.0

Add callback on_before_gw_encode and individual compute_losses for each loss type. Fix bugs

0.3.0

  • Breaking change: remove DeterministGlobaleWorkspace and VariationalGlobalWorkspace in favor of the functions: global_workspace and variational_global_workspace.
  • Allow setting custom GW encoders and decoders.
  • Breaking change: remove self.input_dim, self.encoder_hidden_dim, self.encoder_n_layers, self.decoder_hidden_dim, and self.decoder_n_layers in GWModules.

0.3.1

Fix bugs related to imports and default_decoders.

0.3.2

  • Revert to using classes for GWs (it's easier when loading from checkpoints.)

  • GlobalWorkspace is renamed to GlobalWorkspaceBase and GlobalWorkspace now refers to DeterministicGlobalWorkspace.

0.4.0

  • Use ABC for abstract methods.
  • Replace DomainDescription with GWInterface.
  • Add contrastive_fn attribute in DeterministicGWLosses to compute the contrastive loss. It can then be customized.
  • Rename every abstract class with ClassNameBase. Rename every "Deterministic" classes to remove "Deterministic".
  • Remove all config related functions. This is not the role of this repo.

0.5.0

  • Replace loss coef buffers by a LossCoef TypedDict.
  • Add RepeatedDataset to shimmer.
  • Add docs in docs/, API documentation in https://ruflab.github.io/shimmer/, and some code examples.
  • Replace Black, isort, and flake8 with Ruff (see #8).
  • Remove GWInterfaces entirely and favor giving encoders and decoders directly to the GWModule. See the updated example examples/main_example/train_gw.py to see what changes to make (see #9).
  • Remove GWModuleBase.translate and GWModuleBase.cycle. Translation and cycles can now be done with the utils function translation and cycle.
  • Remove GlobalWorkspaceBase.batch_demi_cycles, GlobalWorkspaceBase.batch_cycles, and GlobalWorkspaceBase.batch_translations. This can be done with utils functions of the same name.
  • Rename GWModuleBase.fusion_mechanism to GWModuleBase.fuse, GWModuleBase.encode to GWModuleBase.encode_and_fuse, and GWModuleBase.encode_pre_fusion to GWModuleBase.encode. Same for the associated methods in GlobalWorkspaceBase.
  • Remove on_before_gw_encode_{loss} callbacks to allow sharing computation between loss functions.
  • Remove many _with_uncertainty functions. The GWModuleWithUncertainty now behaves like the other GWModules.
  • Rename all "with_uncertainty" methods to "bayesian". Note, BayesianGlobalWorkspaces are still a work in progress.
  • Added selection mechanisms (inheriting from SelectionBase, see docs) to fuse representations according to different mechanisms (e.g. Attention).
  • GlobalWorkspace (and associated GWModule, GWLosses, ...) now uses the RandomSelection mechanism. For the old behavior, use GlobalWorkspace2Domains.

0.6.0

  • Allow for some domain modules to be trained end-to-end with the global workspace. This brings some breaking changes:
    1. DomainModule.compute_loss and DomainModule.compute_*_loss now require an 3rd parameter raw_target: Any that stores the raw domain input (before being encoded). This is usefull for unimodal losses that require the actual inputs to compute the loss.
    2. GWLossesBase.step requires a new first argument raw_data: RawDomainGroupsT to pass the raw_targets to the domain modules.