Skip to content

Latest commit

 

History

History
175 lines (152 loc) · 8.33 KB

CHANGELOG.md

File metadata and controls

175 lines (152 loc) · 8.33 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 uses Semantic Versioning.

[1.3.1] - 2023-3-20

Fixed

  • Versions of PyTorch Lightning between 1.6.0 and 1.9.X are now supported

[1.3.0] - 2022-8-20

Added

  • HybridModel and HybridPretrainedModel now take additional optional parameters user_metadata and user_metadata_layers_dims
  • get_data.py now includes get_user_metadata

Changed

  • Added item_metadata_layers_dims and user_metadata_layers_dims parameters to HybridPretrainedModel and HybridModel and removed metadata_layers_dims
  • Updated notebooks and examples to include usage of user_metadata

[1.2.2] - 2022-7-14

Fixed

  • a Value Error is now raised when item_metadata contains nulls

[1.2.1] - 2022-7-12

Fixed

  • Deprecated PyTorch Lightning unit test

[1.2.0] - 2022-1-18

Added

  • option to force_split to stratified_split
  • better type hints for Callables
  • added methods get_user_predictions and user_user_similarity to the BasePipeline
  • added _get_user_embeddings method to all model classes

Changed

  • default Dockerfile image to be torch@1.10.0 with CUDA 11.3
  • check if index is in-bound for get_item_predictions and item_item_similarity before calling the model
  • added enable_model_summary and detect_anomaly parameters to CollieMinimalTrainer and deprecated weights_summary and terminate_on_nan to more closely match the new pytorch_lightning API
  • clarified error message when user has a single interaction when using stratified_split
  • updated all examples, tests, and notebooks with post-1.5.0 PyTorch Lightning APIs

Fixed

  • device error when running metrics for a MultiStagePipeline models
  • CollieMinimalTrainer model summary to work with later versions of PyTorch Lightning

Removed

  • default num_workers for Interactions DataLoaders

[1.1.2] - 2021-8-17

Added

  • string support for Adagrad optimizer in model pipelines

[1.1.1] - 2021-8-17

Changed

  • added property max_epochs to CollieTrainer and CollieMinimalTrainer with setter method
  • CollieTrainer's default max_epochs from 1000 to 10

Fixed

  • used new API for setting verbosity in ModelSummary in CollieMinimalTrainer

[1.1.0] - 2021-7-15

Added

  • multi-stage model template MultiStagePipeline
  • multi-stage models HybridModel and ColdStartModel

Changed

  • optimizers and learning rate schedulers are now reset upon each call of fit in CollieMinimalTrainer, matching the behavior in CollieTrainer
  • HybridPretrainedModel now includes bias terms from the MatrixFactorizationModel in score calculation
  • item_item_similarity now uses a more efficient, on-device calculation for item-item similarity
  • optimizers are now stepped using the optimizer_step method for CollieMinimalTrainer
  • _get_item_embeddings methods now return a torch.tensor type on device
  • _move_external_data_to_device optional method to all models

Removed

  • MultiOptimizer.step method

[1.0.1] - 2021-7-13

Fixed

  • GitHub URL in read_movielens_posters_df to point to new repo name

[1.0.0] - 2021-7-13

Changed

  • name of library to collie!

[0.6.1] - 2021-7-13

Added

  • name change warning from collie_recs -> collie

[0.6.0] - 2021-7-6

Added

  • support for explicit data with ExplicitInteractions and explicit_evaluate_in_batches
  • warnings for invalid adaptive loss vs. num_negative_samples combinations

Changed

  • default Dockerfile image to be torch@1.9.0 with CUDA 10.2

Removed

  • hybrid_matrix_factorization_model.py deprecated filename

[0.5.0] - 2021-6-11

Added

  • new model architectures CollaborativeMetricLearningModel, MLPMatrixFactorizationModel, and DeepFM

Changed

  • filename for HybridPretrainedModel to hybrid_pretrained_matrix_factorization.py. The former model filepath is now deprecated and will be removed in future version 0.6.0
  • collie.model.base is now split into its own directory with the same name
  • reduced boilerplate docstrings required for models
  • all model.freeze() -> model.eval()
  • bumped version of sphinx-rtd-theme to 0.5.2

[0.4.0] - 2021-5-13

Added

  • CollieMinimalTrainer for a faster, simpler version of CollieTrainer
  • remove_duplicate_user_item_pairs argument to Interactions

Changed

  • renamed BasePipeline.hparams.n_epochs_completed_ -> BasePipeline.hparams.num_epochs_completed

Fixed

  • a proper ValueError is now raised if no train data is passed into a model
  • loss docstrings that incorrectly stated **kwargs would be accepted

[0.3.0] - 2021-5-10

Changed

  • disable automated batching in ApproximateNegativeSamplingInteractionsDataLoader and HDF5InteractionsDataLoader

[0.2.0] - 2021-4-28

Changed

  • convert_to_implicit will now remove duplicate user/item pairs in DataFrame

[0.1.4] - 2021-4-27

Fixed

  • duplicate user/item pairs in Interactions are now dropped from the COO matrix during instantiation

[0.1.3] - 2021-4-26

Added

  • ability to run stratified_split without any joblib.Parallel parallelization
  • data quality checks to Interactions.__init__ to assert users and items and mat are not None and ratings does not contain any 0s (if so, those rows will now automatically be filtered out)
  • increased test coverage
  • header table to all Jupyter notebooks with links to Colab and GitHub

Changed

  • default processes for stratified_split is now -1
  • default k value in mapk is now set to 10
  • when GPU is available but not set, CollieTrainer now sets it to 1 rather than -1
  • all models now check that train_loader and val_loader attributes are consistent during initialization
  • default unseen_items_only in BasePipeline.get_item_predictions method is now False
  • docs in get_recommendation_visualizations to be clearer
  • get_recommendation_visualizations data quality checks have been moved to the beginning of the function to potentially fail faster
  • create_ratings_matrix no longer raises ValueError if users and items do not start at 0
  • refactored adaptive_hinge_loss

Removed

  • kwargs option for methods that did not explicitly need them

Fixed

  • typo in cross_validation.py error message
  • head and tail methods in interactions/datasets.py to no longer error with n < 1 or large n values
  • num_users and num_items are no longer incorrectly incremented when meta key is provided in HDF5Interactions
  • type hints for device now also include instances of torch.device
  • the type of metadata tensors sent to HybridPretrainedModel are now consistent across all input options
  • removed ineffective quality checks in HybridPretrainedModel.save_model
  • no longer use deprecated nn.sigmoid in library
  • a relu final activation layer now works in NeuralCollaborativeFiltering model
  • df_to_html now outputs proper HTML when multiple html_tags options are specified
  • tutorial notebooks now fully run on Colab without having to only rely on previously-saved data
  • add value of 1e-11 to BasePipeline.get_item_predictions denominator to avoid potential NaNs

[0.1.2] - 2021-4-14

Added

  • various badges to README
  • links to documentation where relevant
  • Colab links to tutorial notebooks and README quickstart

Changed

  • ApproximateNegativeSamplingInteractionsDataLoader uses a sampler instead of a batch_sampler for greater speed increases
  • base Dockerfile image to the torch@1.8.1 version
  • read_movielens_posters_df now works as expected even when data/movielens_posters.csv file does not exist
  • renamed LICENSE.txt -> LICENSE
  • renamed pull_request_template.md -> PULL_REQUEST_TEMPLATE.md

[0.1.1] - 2021-4-13

Added

  • GitHub Actions and templates in .github

[0.1.0] - 2021-4-13

Added

  • Collie library for open sourcing

[0.0.0] - 2021-4-12

Added

  • Initial project scaffolding