Fix normalization in translation models
Improve loading function of wikidatavitals datasets.
Fix ConvKB scoring function and normalization step
Fix the documentation in evaluation and inference modules
Fix a typo in the sampling module’s documentation
Add support of Python 3.7 back
Add relation prediction evaluation
Add relation negative sampling module
Add inference module
Update models’ API accordingly to the previous new features
Switch from TravisCI to GitHub Actions
Update in available pretrained models
Removed useless k_max parameter in link-prediction evaluation method
Add pretrained version of TransE for yago310 and ComplEx for fb15k237 and wdv5.
Add pretrained version of TransE for Wikidata-Vitals level 5
Add support for Python 3.8
Clean up loading process for kgs
Fix deprecation warning
Add data loader for wikidata vitals knowledge graphs
Bug fix get_ranks method
Bug fix in KG split method
Fix WikiDataSets loader (again)
Fix WikiDataSets loader
Fix reduction in BCE loss
Add pretrained models
Fix bug in pre-trained models loading that made all models being redownloaded every time
Minor bug patch
Update urls to retrieve datasets and pre-trained models.
Add binary cross-entropy loss
Change API for pre-trained models
Patch in pre-trained model loading
Added pre-trained loading for TransE on FB15k237 in dimension 100.
Add parameter in data redundancy to exclude know reverse triplets from duplicate search.
Add methods to compute data redundancy in knowledge graphs as in 2020 paper by Akrami et al (see references in concerned methods).
Patch an awkward import
Add dataset loaders for WN18RR and YAGO3-10
Redefinition of the models’ API (simplified interfaces, renamed LP methods and added get_embeddings method)
Implementation of the new API for all models
TorusE implementation fixed
TransD reimplementation to avoid matmul usage (costly in back-propagation)
Added feature to negative samplers to generate several negative samples from each fact. Those can be fed directly to the models.
Added some wrappers for training to utils module.
Progress bars now make the most of tqdm’s possibilities
Defined a new homemade and simpler DataLoader class.
Removed the use of torch DataLoader object.
Added a method to print results in link prediction evaluator
Fixed a misfit test
Cleared the definition of rank in link prediction
Improved use of tqdm progress bars
Change in the API of loss functions (margin and logistic loss)
Added ConvKB model
Minor patch in interfaces
Various bug fixes
New KG splitting method enforcing all entities and relations to appear at least once in the training set.
Minor bug fixes
Minor bug fixes
Fixed requirements conflicts
Added TorusE model
Fixed some bugs
Fixed error in bilinear models.
Added intermediate function for hit@k metric in link prediction.
Fixed assertion error in Analogy model
Implemented Triplet Classification evaluation method
Added Negative Sampler objects to standardize negative sampling methods.
Implemented HolE model (Nickel et al.)
Implemented ComplEx model (Trouillon et al.)
Implemented ANALOGY model (Liu et al.)
Added knowledge graph splitting into train, validation and test instead of just train and test.
Implemented Bernoulli negative sampling as in Wang et al. paper on TransH (2014).
Implemented Mean Reciprocal Rank measure of performance.
Implemented Logistic Loss.
Changed implementation of margin loss to use torch methods.
Changed implementation of LinkPrediction ranks by moving functions to model methods.
Fixed a major bug/problem in the Evaluation protocol of LinkPrediction.
Minor bug fixes in the various normalization functions.
Fixed CUDA support.
Added support for filtered performance measures.
First real release on PyPI.
First release on PyPI.