Roadmap for v0.60 - Githubissues

Alcoholrithm / TabularS3L

A PyTorch Lightning-based library for self- and semi-supervised learning on tabular data.

MIT License

26 stars 2 forks source link

Roadmap for v0.60 #18

Open Alcoholrithm opened 1 month ago

Alcoholrithm commented 1 month ago

In the upcoming version of the package, I plan to introduce the following updates and features:

1. Support for Embedding Modules :white_check_mark:

[x] Identity
[x] Feature Tokenizer from FT-Transformer
[x] Introduce EmbeddingConfig

2. Support for Diverse Backbones :white_check_mark:

[x] MLP
[x] Transformer
[x] Introduce BackboneConfig

3. Model Updates to Support New Features :white_check_mark:

[x] DAE
[x] VIME
[x] SubTab
[x] SCARF
[x] SwitchTab

4. Split Reconstruction Head

[ ] Implement separate reconstruction heads for categorical and continuous features, enhancing the handling of heterogeneous data.

5. Update Test Files and Benchmark Code :white_check_mark:

[x] Test Files
[x] Benchmark Code

6. Code Cleanup and Refactoring :white_check_mark:

[x] Refactoring
[x] Code Cleanup