Issues

• [BUG] fit() crashes with BatchNorm when the training rows leave a one-row last batch (e.g. 162 rows)

  bug
  OpenTabular/DeepTab#577 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] load() across the LSS boundary builds a broken estimator (e.g. MLPRegressor.load on an LSS artifact)

  bug
  OpenTabular/DeepTab#567 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] LSS continuation fits crash when family options contain a NumPy array or tensor (e.g. quantiles=np.linspace(...))

  bug
  OpenTabular/DeepTab#563 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] HPO searches ResNet's boolean norm over layer names, so norm=False is never tried

  bug
  OpenTabular/DeepTab#559 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] Distribution families and fitted LSS estimators cannot be pickled (local lambdas in BaseDistribution)

  bug
  OpenTabular/DeepTab#557 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] LSS save() returns None, requires a path and emits no save events, unlike the other estimators

  bug
  OpenTabular/DeepTab#547 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] save(pathlib.Path) fails and writes no artifact with log_to_file=True and verbosity=3

  bug
  OpenTabular/DeepTab#545 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] ObservabilityConfig(mlflow_log_model=False) is ignored: checkpoints are still uploaded to MLflow

  bug
  OpenTabular/DeepTab#543 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] build_model() on a fitted estimator keeps is_fitted_=True with an untrained network

  bug
  OpenTabular/DeepTab#541 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] predict()/predict_proba() crash after fit(precision="bf16-mixed"): unsupported ScalarType BFloat16

  bug
  OpenTabular/DeepTab#539 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] Contrastive pretraining crashes on a batch too small to form pairs (e.g. a one-row last batch)

  bug
  OpenTabular/DeepTab#529 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] Classifier contrastive pretraining crashes when a batch contains a class with a single row

  bug
  OpenTabular/DeepTab#527 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] Regressor score() reloads the training checkpoint from disk: fails when it is gone and overwrites the current weights

  bug
  OpenTabular/DeepTab#503 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] LSS estimators ignore ObservabilityConfig: no lifecycle events, run directory or experiment-tracker logging

  bug
  OpenTabular/DeepTab#499 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] fit(enable_model_summary=False) is ignored because a ModelSummary callback is always added

  bug
  OpenTabular/DeepTab#485 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] Sequence architectures count tokens without the CLS and external-embedding tokens (shuffle_embeddings drops tokens, learned_flatten crashes)

  bug
  OpenTabular/DeepTab#483 · Filter by author ChrisW09 opened on Oct 21, 2026

• [BUG] pooling_method="rnn" and "conv" crash at the first forward pass on every pooling architecture

  bug
  OpenTabular/DeepTab#481 · Filter by author ChrisW09 opened on Oct 21, 2026

• [FEATURE] ROCm (AMD GPU) support

  enhancement
  OpenTabular/DeepTab#479 · Filter by author Svenum opened on Oct 18, 2026
  1

• Add dedicated foundation-model test job

  enhancement
  OpenTabular/DeepTab#393 · Filter by author mkumar73 opened on Jun 7, 2026 · M7 - Model Expansion
  

• Add optional extras for foundation model dependencies

  enhancement
  OpenTabular/DeepTab#392 · Filter by author mkumar73 opened on Jun 7, 2026 · M7 - Model Expansion
  

• Add TabICL adapter

  enhancement
  OpenTabular/DeepTab#391 · Filter by author mkumar73 opened on Jun 7, 2026 · M7 - Model Expansion
  

• Add TabPFN adapter

  enhancement
  OpenTabular/DeepTab#390 · Filter by author mkumar73 opened on Jun 7, 2026 · M7 - Model Expansion
  

• Define foundation-model adapter interface

  enhancement
  OpenTabular/DeepTab#389 · Filter by author mkumar73 opened on Jun 7, 2026 · M7 - Model Expansion
  

• [FEATURE] Potential optimization for selective_scan_seq (einsum discretisation)

  enhancement
  OpenTabular/DeepTab#250 · Filter by author nwuestefeld opened on Mar 20, 2025

• [FEATURE] Embedding layer: fuse pre-encoded column-name embeddings with value embeddings

  enhancement
  OpenTabular/DeepTab#227 · Filter by author AnFreTh opened on Feb 19, 2025 · M8 - Optimization and deployment