Repository Issue Activity (beta)

#1622 [Models] Support historical-only exogenous variables in DeepAR

#1619 Relax pytorch-lightning <2.6.0 cap: blocks the CVE-2026-58659 fix in 2.6.6

#1615 DeepAR: decoder_hidden_layers>0 with default decoder_hidden_size=0 silently builds zero-width hidden layers

#1608 Most models do not allow user to specify activation of the final layer?

#1616 xLSTM: decoder_activation is neither validated nor warned about when decoder_layers=1

#1614 AutoSOFTSSharp with Ray backend fails with CUDNN_STATUS_SUBLIBRARY_VERSION_MISMATCH on NVIDIA A800 CUDA 13.0

#1098 Categorical Features for TFT

#1601 [Losses] _divide_no_nan turns an infinite loss into 0.0, so broken trials win Auto search

#1595 [QUESTION] How should I choose `max_steps` when comparing with epoch-based training?

#1304 Python 3.12: SyntaxWarning: invalid escape sequence

#1578 AttributeError: 'NeuralForecast' object has no attribute 'static_scalers_'

#1045 [core] Can't use DBFS as a filesystem in distributed

#1569 [FEAT] Categorical feature support for multivariate models

#1291 Auto models get ray.exceptions.ActorDiedError when run on multi-GPU node

#1558 Validation loss computation in multi-device training

#559 Ray error with AutoHINT

#1037 [common] Issue with multi-gpu and `ddp_spawn` strategy when running predict

#686 Code stuck on "initalizing ddp" when using more than one gpu on neuralforecast AutoTFT, AutoNHITs

#494 Error when using AutoTunning

#1310 Significant gap in forecasts from v2.0.1 to v3.0.0 when using DistributionLoss

#707 Neuralforecast different predictions

#916 [Core] Categorical Features

#1301 Recommend a versatile and easy-to-implement loss function

#1100 All Multivariate Models (Except SOFTS) cannot be trained on multiple GPUs

#571 Default configs of AutoNHITS seems invalid