v1.0.5

Bug Fixes

• remove dead code for plots (f96795f)

Features

• switch to sparse matrix trick (98910bc)

v1.0.4

Bug Fixes

• allow smaller different nshared and nindependent (4b365a7)
• sparsemax on train and predict epoch (6f7c0e0)

v1.0.3

Bug Fixes

• map class predictions for XGB results (3747e2f)

Features

• fix shared layers with independent batchnorm (5f0e43f)

v1.0.2

Bug Fixes

• multiclass prediction mapper (2317c5c)
• remove deepcopy from shared blocks (123932a)

v1.0.1

Bug Fixes

• regression: fix scheduler (01e46b7)
• fixing Dockerfile for poetry 1.0 (6c5cdec)
• importance indexing fixed (a8382c3)
• local explain all batches (91461fb)
• regression gpu integration an typos (269b4c5)
• resolve timer issue and warnings (ecd2cd9)

Features

• improve verbosity (8a2cd87)

v1.0.0

Bug Fixes

• deps: update dependency numpy to v1.17.3 (eff6555)
• deps: update dependency numpy to v1.17.4 (a80cf29)
• deps: update dependency torch to v1.3.1 (18ec79b)
• deps: update dependency tqdm to v4.37.0 (f8f04e7)
• deps: update dependency tqdm to v4.38.0 (0bf45d2)
• functional balanced version (fab7f16)
• remove torch warnings (index should be bool) (f5817cf)

Features

• add gpu dockerfile and adapt makefile (8d14406)
• update notebooks for new model format (43e2693)

v0.1.2

Bug Fixes

• add softmax to predict_proba (bea966f)
• correct code linting (ae3098c)
• float type when output_dim=1 (7bb7dfd)

Features

• add editorconfig file (5e84b66)
• add flake8 (b1be1f9)
• run flake8 in CI (e72d416)
• start PyTorch TabNet Paper Implementation (e7dc059)