dask / dask/dask-ml

Tests failing with `ValueError: cannot broadcast shape (nan,) to shape (nan,)`

Open
#1,012 4 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
951
Forks
262
PR merge metrics
No merged PRs in 30d

Description

**Describe the issue**:

When updating `dask` to 2025.2.0, the following `dask-ml` tests start failing:
```
=========================== short test summary info ============================
FAILED tests/ensemble/test_blockwise.py::TestBlockwiseVotingRegressor::test_fit_array - ValueError: cannot broadcast shape (nan,) to shape (nan,)
FAILED tests/ensemble/test_blockwise.py::TestBlockwiseVotingRegressor::test_fit_frame - ValueError: cannot broadcast shape (nan,) to shape (nan,)
FAILED tests/linear_model/test_glm.py::test_lr_score - ValueError: cannot broadcast shape (nan,) to shape (nan,)
FAILED tests/metrics/test_regression.py::test_ok[r2_score-True] - ValueError: cannot broadcast shape (nan,) to shape (nan,)
FAILED tests/metrics/test_regression.py::test_ok[r2_score-False] - ValueError: cannot broadcast shape (nan,) to shape (nan,)
FAILED tests/preprocessing/test_data.py::TestQuantileTransformer::test_fit_transform_frame - IndexError: boolean index shape (nan,) must match indexed array's (nan,).
FAILED tests/test_incremental.py::test_scoring_string[threads-r2] - ValueError: cannot broadcast shape (nan,) to shape (nan,)
FAILED tests/test_incremental.py::test_scoring_string[distributed-r2] - ValueError: cannot broadcast shape (nan,) to shape (nan,)
FAILED tests/test_parallel_post_fit.py::test_laziness - ValueError: cannot broadcast shape (nan,) to shape (nan,)
= 9 failed, 585 passed, 116 skipped, 1 deselected, 8 xfailed, 6 xpassed, 312 warnings in 204.79s (0:03:24) =
```

**Minimal Complete Verifiable Example**:

```console
$ pytest
```

**Anything else we need to know?**:

Context: bumping dask in nixpkgs: https://github.com/NixOS/nixpkgs/pull/381896

**Environment**:

- Dask version: 2025.2.0
- Python version: 3.12.8
- Operating System: NixOS
- Install method (conda, pip, source): nix

Contributor guide

Open the contributing guide

Research direction

Start by running pytest with Dask 2025.2.0 and inspect the failures in tests/ensemble/test_blockwise.py, tests/linear_model/test_glm.py, tests/metrics/test_regression.py, tests/preprocessing/test_data.py, tests/test_incremental.py, and tests/test_parallel_post_fit.py. Trace the broadcast and boolean-index shape errors to determine the compatibility change, then verify that the full suite passes without regressing the reported tests.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning, testing
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
32/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.