All pull requests

Pull requests

• fix(evaluation): validate univariate inputs in get_metrics_per_ts and calculate_seasonal_error

  #3369·Hrushikesh-ramilla opened Nov 7, 2026·

  ·
  1

• feat(time_feature): support sub-second frequencies

  #3368·Hrushikesh-ramilla opened Nov 6, 2026·

  ·
  1

• fix(dataset): normalize period frequencies in PandasDataset

  #3367·Hrushikesh-ramilla opened Nov 6, 2026·

  ·
  1

• fix(dataset): support TestData and rolling windows in MultivariateGrouper

  #3366·Hrushikesh-ramilla opened Nov 6, 2026·

  ·
  1

• Clamp max_history to available observations in splitter

  #3365·Hrushikesh-ramilla opened Nov 6, 2026·

  ·
  1

• fix(time_feature): filter seasonal lags consistently with num_default_lags

  #3364·Hrushikesh-ramilla opened Nov 5, 2026·

  ·
  1

• Cap requested test history at the available observations

  #3362·student-shivamm opened Nov 1, 2026·

  ·
  1

• Support pandas 3.0.x

  #3360·celestinoxp opened Oct 20, 2026·

  ·
  2

• Torch: allow enable_checkpointing=False in trainer_kwargs

  #3359·Thanatos9404 opened Oct 19, 2026·

  ·
  1

• Create BinnedUniforms samples on the parameters' device and dtype

  #3356·raashish1601 opened Oct 7, 2026·

  ·

• Fix BinnedUniformsOutput construction and distribution arguments

  #3355·raashish1601 opened Oct 7, 2026·

  ·

• Fix BinnedUniforms.mean dividing by the number of bins

  #3354·raashish1601 opened Oct 7, 2026·

  ·

• fix: reject mismatched forecast counts during evaluation

  #3353·rioyu123 opened Oct 3, 2026·

  ·
  1

• Fix ZeroDivisionError in get_seasonality for zero-multiple frequencies

  #3352·mayuriphad opened Sep 30, 2026·

  ·
  1

  3

• Keep the predictor on CPU when training with accelerator="cpu"

  #3351·ankitdhandharia opened Sep 29, 2026·

  ·

• Fix Uber TLC series starting at the wrong date

  #3350·David-Wu1119 opened Sep 24, 2026·

  ·
  1

• fix(nursery): @validted -> @validated — importing pts.model.n_beats raises NameError

  #3347·Anai-Guo opened Sep 9, 2026·

  ·

• fix(torch): narrow training-recovery catch and preserve the traceback (#3295)

  #3346·warren830 opened Sep 4, 2026·

  ·
  1

  1

• fix(torch): handle loc=None in PiecewiseLinearOutput.distribution

  #3345·warren830 opened Aug 30, 2026·

  ·
  1

  2

• fix: handle zero frequency in get_seasonality

  #3344·kluacisco888 opened Aug 30, 2026·

  ·
  1

• Do not mutate the caller's dataframe in PandasDataset

  #3342·VenishPaneliya opened Aug 26, 2026·

  ·

• fix: process past dynamic categorical features

  #3340·mameikagou opened Aug 21, 2026·

  ·

• Fix SampleForecast quantile interpolation

  #3339·tandede opened Aug 21, 2026·

  ·
  1

• Allow toolz 1.0.0

  #3338·MaximGovorchuk opened Aug 9, 2026·

  ·
  1

• Fix weighted_average reduction for dim=0

  #3337·Robertboy18 opened Aug 7, 2026·

  ·
  1