first commit towards adding the TabTransformer. This will be a mix between a number of implementation (that will of course be credited in the code)

added inverse_transform for DeepPreprocessor. Also added NameError and AttributeError to avoid bare excepts

Added comment for Mac users regarding the dataloaders not running in parallel for the latest torch version or python versions >= 3.8. Also, upgraded sub-version of the package to 0.4.7

updated docs so they are consistent with new functionalities. Updated logo. Updated README and fix a typo in setup.py

all started by trying to add a line so that the builder accepted 2D images (or in general images of dim different than 3. But it ended up by adding functionalities so that each individual component (wide, deepdense, deeptext and deepimage) can be used individually

Added a condition to account for 2D images (images with no color channels). This way we replicate exactly what ToTensor does within the WideDeepDataset class

Modified documentation and added doc test. Also changed some of the code according to the increase of test coverage

Modified documentation and added doc test. Also changed some of the code in the main WideDeep class to increase test coverage

Finished implementation of DeepDenseResnet. Adapated DeepDense code for consistency (simply renamed 'embed_p' to 'embed_dropout'). Adapted docs and README. Included unit test for DeepDenseResnet.

Changed loss value printed on the screen for regression. From RMSE to MSE. Adjusted documentation. Also changed unit test for callbacks to the new Linear-Embedding implementation, which implied incresing delta for early stopping

#18 Implementation of the Linear model that is the wide component via an Embedding layer. This helps optimize speed and memory usage. Adapted all the submodules accordingly

Changed README, wide_deep.py, docs, and examples according to the new metrics. Only thing left is adjust the notebooks

fixed a bug related to the focal loss. Remove activations before the losses in the fit method in WideDeep. Sigmoid and logSoftmax happen within the loss. Added activations in the train and eval step and in the predict method

fixing bug related to the Focal Loss. Replaced binary_cross_entropy_with logits with binary_cross_entropy for the binary classification case

fixing bug related to the Focal Loss. Replaced binary_cross_entropy_with logits with binary_cross_entropy

Adapted the examples to the small code changes. Replace add_ with add in WideDeep to avoid annoying warnings. Replace output_dim with pred_dim in Wide for consistentcy

finished docs apart from examples. Changed name of DeepPreprocessor to DensePreprocessor for consistency. Added raise error if model components do not have an 'output_dim' attribute