twcrps_ensemble bounds ignore the input precision
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Newbie friendliness
- 82/100
Research direction
The issue identifies uv_weighted_score_chain in core/utils.py as the cause and says the kernel scores already use asarray_matching for the bounds. Read that helper and compare its bound conversion with the kernel score conversions. Reproduce the float32 NumPy and float64 torch cases from the issue; done means outputs preserve the input precision and the torch result agrees with NumPy.
Written by the indexing model from the issue text.
Description
The threshold bounds a and b of twcrps_ensemble are converted with the default dtype instead of the input's, so float32 numpy inputs return float64 (also with the default bounds), and on torch float64 inputs the bounds are rounded to float32.
obs = np.float32([0.3, 1.2])
fct = np.random.default_rng(0).normal(size=(2, 5)).astype(np.float32)
sr.twcrps_ensemble(obs, fct).dtype # float64
With float64 torch tensors and a=0.1, b=0.7, the result differs from numpy by about 5e-9. The cause is uv_weighted_score_chain in core/utils.py; the kernel scores already convert the bounds with asarray_matching.
- Dominant language
- Python
- Stars
- 108
- Forks
- 15
- Avg merge
- 1h 42m
- Merged PRs (30d)
- 9
Getting set up
- No Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from frazane/scoringrules
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
frazane/scoringrules#164 ·
Maintainers usually reply within 1 day
-
backend:torch bug documentation
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
frazane/scoringrules#146 ·
Maintainers usually reply within 1 day
-
Bugs in the legacy TorchBackendPossibly taken @frazane claimed this today. Openbackend:torch bug
Difficulty 3/5 1-2 days Newbie friendliness 25/100
frazane/scoringrules#147 ·
Maintainers usually reply within 1 day
-
bug
Difficulty 3/5 1-2 days Newbie friendliness 58/100
frazane/scoringrules#145 ·
Maintainers usually reply within 1 day
-
weighted_interval_score adds w_median * median instead of w_median * |obs - median| on all non-numba backendsPossibly taken @lucaluo925 claimed this 39 days ago. Openbug
Difficulty 2/5 1-3 hours Newbie friendliness 35/100
frazane/scoringrules#140 ·
Maintainers usually reply within 1 day
All issues in frazane/scoringrules
Similar issues
-
ChatBedrockConverse changes the keys of JSON tool results to camelCase (`order_id` becomes `orderId`)Possibly taken @geyazheng claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
langchain-ai/langchain-aws#1469 ·
Maintainers usually reply within 1 day
-
Difficulty 1/5 Under an hour Newbie friendliness 88/100
Maintainers usually reply within 1 day
-
b4b usability
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
Maintainers usually reply within 1 day
-
fix(schedule): align list_tasks since cursor filter boundary from strict > to >=Possibly taken @bfs2021 claimed this today. Openagent-ready bug p2 schedule
Difficulty 2/5 1-3 hours Newbie friendliness 83/100
allenporter/supernote#540 ·
Maintainers usually reply within 1 day
-
The docs build fails: ecc's live inventory dropped the ecc.ecc names btclib's pinned ecc usesPossibly taken @fametrano claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
btclib-org/btclib#2968 ·
Maintainers usually reply within 1 day