Nixtla/mlforecast - CodSpeed

mlforecast

Blog Docs Changelog

Performance History

Latest Results

test: remove test for deleted test_weight validation

feat/transfer_cp

2 days ago

Build partition context frames with narwhals Replace the pandas/polars isinstance branching in _update_partition_assignments with narwhals selects and a horizontal concat. The id column now comes from static_features_ (always present, one row per series in uid order), removing the pd.Index/pl.Series special-casing of self.uids, and key-column dtypes are preserved end-to-end so the bucket join sees the same dtypes as at fit time. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

simonez-tuidi:feature/pooled_partition_by

3 days ago

Merge branch 'main' into feat/transfer_cp

feat/transfer_cp

4 days ago

(fix) mypy errors

feat/transfer_cp

4 days ago

fix(lint): remove unused dataclasses.field import

feat/transfer_cp

4 days ago

fix(transfer_cp): correct sign in compute_conformity_scores signed mode (y-pred, not pred-y) Fixed sign inversion bug in signed conformity scores: now correctly computes y - pred (residuals where positive = underprediction, negative = overprediction). Updated test assertions and docstring accordingly. This ensures prediction intervals are correctly positioned relative to forecasts. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

feat/transfer_cp

4 days ago

Document pooled slow-path and memory limitations Quantile and seasonal-rolling transforms have no aggregate fast path in pooled modes; note the cost in their docstrings and the pooled guide, along with the pooled-state memory behavior (full history kept regardless of keep_last_n, per-model backup at predict). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

simonez-tuidi:feature/pooled_partition_by

5 days ago

Merge branch 'feature/pooled_partition_by' of github.com:simonez-tuidi/mlforecast into feature/pooled_partition_by

simonez-tuidi:feature/pooled_partition_by

5 days ago

Latest Branches

-1%

[FEAT] Transfer learning with Conformal Prediction#628

2 days ago

1514b2a

feat/transfer_cp

0%

Adding `partition_by` to lag-transforms building upon PooledState#651

3 days ago

db83253

simonez-tuidi:feature/pooled_partition_by

0%

[FIX] Treat null/NaN keys as equal in pooled transforms#658

5 days ago

284590d

simonez-tuidi:fix/nulls-handling-in-pooled-transforms

© 2026 CodSpeed Technology

Home Terms Privacy Docs