v3.2.0

[Marco (marcopeix)]

New features

• [FEAT] Support categorical exog for univariate models Marco (@marcopeix) ([FEAT] Support categorical exog for univariate models #1552)
• [FEAT] Add FreDF loss combining time-domain MSE and frequency-domain MAE Dresden (@DresdenGman) (add FreDF loss combining time-domain MSE and frequency-domain MAE #1563)
• [FEAT] Add GELU as a grn_activation option in TFT Kai Wöllstein (@autokaiai) ([FEAT] Add GELU as a grn_activation option in TFT #1560)
• [FEAT] Remove cpus and gpus in favor of ray_options Marco (@marcopeix) ([FEAT] Remove cpus and gpus in favor of ray_options #1544)

Bugfixes

• [FIX] Multi-device training for multivariate models Marco (@marcopeix) ([FIX] Multi-device training for multivariate models #1557)
• [FIX] Translate input_size_multiplier in user-supplied Auto configs Vedant Agarwal (@Vedant-Agarwal) (Translate input_size_multiplier in user-supplied Auto configs #1553)
• [FIX] Further fixes for CV with no df Marco (@marcopeix) ([FIX] Further fixes for CV with no df #1548)
• [FIX] Allow CV without passing a df Marco (@marcopeix) ([FIX] Allow CV without passing a df #1545)
• [FIX] Prevent a second optimization study when using conformal predictions Marco (@marcopeix) ([FIX] Prevent a second optimization study when using conformal predictions #1539)
• [FIX] Raise error when missing params in auto config Marco (@marcopeix) ([FIX] Raise error when missing params in auto config #1540)

Documentation

• [DOC] Add xLSTM to docs Marco (@marcopeix) ([DOC] Add xLSTM to docs #1543)
• [DOC] Add tutorial to convert nf models to onnx Marco (@marcopeix) ([DOC] Add tutorial to convert nf models to onnx #1542)

---

This discussion was created from the release v3.2.0.