Collie

Latest version: v1.3.1

Safety actively analyzes 693883 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 2 of 4

1.1.1

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

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

Fixed
- GitHub URL in ``read_movielens_posters_df`` to point to new repo name

1.0.0

Changed
- name of library to ``collie``!

0.6.1

Added
- name change warning from ``collie_recs -> collie``

0.6.0

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 ``torch1.9.0`` with CUDA 10.2
Removed
- ``hybrid_matrix_factorization_model.py`` deprecated filename

Page 2 of 4

© 2025 Safety CLI Cybersecurity Inc. All Rights Reserved.