Embeddings

[AnFreTh]

This pull request includes significant updates to the testing workflow and various model classes to enhance functionality and maintainability. The most important changes involve adding a new GitHub Actions workflow for unit tests, removing the rotary embedding from the attention mechanism, and updating the embedding layer to handle additional embedding features.

Testing Workflow:

• .github/workflows/pr-tests.yml: Added a new GitHub Actions workflow to run unit tests on pull requests targeting the develop and master branches. This workflow sets up Python, installs dependencies, runs tests using pytest, and enforces test success before allowing merges.

Model Class Updates:

• mambular/arch_utils/layer_utils/attention_utils.py: Removed the rotary parameter and associated logic from the Attention and Transformer classes, simplifying the attention mechanism. [1] [2] [3] [4] [5]

Embedding Layer Enhancements:

• mambular/arch_utils/layer_utils/embedding_layer.py: Enhanced the EmbeddingLayer class to handle an additional emb_feature_info parameter and updated the forward method to process these embedding features. [1] [2] [3] [4] [5] [6] [7]

Base Model Adjustments:

• mambular/base_models/basemodel.py: Modified the encode method to accept a single data parameter instead of separate num_features and cat_features parameters. [1] [2]

FTTransformer and Lightning Wrapper Updates:

• mambular/base_models/ft_transformer.py: Updated the FTTransformer class to accept a tuple of feature information and adjusted the forward method accordingly. [1] [2] [3]
• mambular/base_models/lightning_wrapper.py: Refactored the LightningWrapper class to handle the new feature information tuple and updated various methods to align with this change. [1] [2] [3] [4] [5] [6] [7] [8] [9]