The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Error code: FileFormatMismatchBetweenSplitsError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
UQE2E: Fine-tuned weights and evaluation data
Companion data for the manuscript
Uncertainty-Aware End-to-End AI Weather Forecasting: Disentangling Observation and Model Contributions Rodrigo Almeida, Noelia Otero, Jost Arndt, Simon Baur, Wojciech Samek, Jackie Ma Applied Machine Learning Group, Fraunhofer Heinrich-Hertz Institute, Berlin
Preprint: arxiv.org/abs/2608.30795 Code: https://gitlab.hhi.fraunhofer.de/ai-aml/uqe2e.
This repository holds:
- the fine-tuned model runs (selected checkpoint per run and per lead): two probabilistic encoders, two MC-dropout processors, the control processor and the three joint end-to-end fine-tunes,
- the normalisation factors as used by the runs,
- a few small static auxiliary arrays the data loaders expect,
- the evaluation caches behind every figure and number of the manuscript and its reviewer reply.
Everything else (raw observations, upstream deterministic weights, ERA5,
climatology, WeatherBench 2 scores) is either in the upstream Aardvark
release or regenerable with the code; see the code repository's DATA.md.
Layout
The tree mirrors the code repository, so a single download into the repository root puts every file where the code looks for it:
uv run hf download rodrigoalmeida1994/uqe2e --repo-type dataset --local-dir .
data/
outputs/
encoder_noise_cond_embed_ft_full_fixeddata00z/encoder_deviousiguana/ # prob. encoder, embedding noise (epoch 0)
encoder_noise_cond_embed_ft_full_fixeddata00z_normmode/encoder_glaringsawfly/ # prob. encoder, norm noise (epoch 10)
processor_mc_dropout_crps_fixeddata00z_7member/processor_malachitegoshawk/ # MC-dropout processor paired with deviousiguana
processor_mc_dropout_crps_fixeddata00z_normmode_7member/processor_powerfulzebra/ # MC-dropout processor paired with glaringsawfly
processor_mc_dropout_crps_fixeddata00z_detic_7member/processor_enigmaticmalkoha/ # control processor (deterministic analyses)
e2e_prob_multilead00z_lead0_gridded_normmode/tas/e2e_cobaltmillipede/ # joint fine-tune, gridded objective (epoch 17)
e2e_prob_multilead00z_lead0_gentle_normmode/tas/e2e_xanthicmarmot/ # joint fine-tune, station objective, T2M (epoch 12)
e2e_prob_multilead00z_lead0_gentle_normmode/ws/e2e_excellentpetrel/ # joint fine-tune, station objective, WS10 (epoch 17)
norm_factors/ # mean_*/std_* per stream, hirs_means/hirs_stds
era5/ # elev_vars_1.npy, era5_orog_0p25deg_m.npz, era5_pressure_levels_4u*.npy
lat_weights/weights_lat_1.npy
era5_spatial_means.npy era5_avdiff_means.npy era5_avdiff_stds.npy loss_weights.npy
hadisd_station_info_v343_2025f.txt
results/ # evaluation caches read by uq_e2e.paper_figures
Runs
Each run directory holds config.pkl, the selected checkpoint epoch_<k>
and the per-epoch validation logs (losses_*.npy, rmse_*.npy,
train_losses_*.npy, val_metrics_history_0.json) from which the
checkpoint was selected (uq_e2e.checkpoint_selection.select_best_epoch).
Processor runs have one forecast_<lead>/ sub-directory per daily rollout
step (leads 1..10), each with its own selected checkpoint. Joint fine-tunes
hold the whole pipeline (encoder, per-lead processors and, for the station
objective, the per-lead station decoders) in one model_state_dict, plus
lead_weights.json; load them with uq_e2e.e2e_model.load_finetuned_e2e.
Checkpoints of non-selected epochs and the push-forward intermediates
(ic_*.mmap) are not included.
| Run | Role | Selected epoch(s) |
|---|---|---|
encoder_deviousiguana |
probabilistic encoder, embedding noise mode, 7-member fair-CRPS fine-tune of the upstream encoder |
0 |
encoder_glaringsawfly |
probabilistic encoder, norm (conditional LayerNorm) noise mode; headline encoder |
10 |
processor_malachitegoshawk |
processor with trained-in MC dropout (p=0.05, fair CRPS), push-forward on deviousiguana's analyses | leads 1..10: 8, 1, 0, 0, 0, 1, 0, 0, 0, 0 |
processor_powerfulzebra |
same, on glaringsawfly's analyses; headline component pair | leads 1..10: 9, 1, 0, 0, 0, 1, 1, 0, 0, 0 |
processor_enigmaticmalkoha |
same recipe on the deterministic upstream encoder's analyses: the compensation control (its dropout branch never saw encoder noise) | leads 1..10: 8, 1, 1, 0, 0, 1, 1, 1, 0, 1 |
e2e_cobaltmillipede |
joint fine-tune of glaringsawfly + powerfulzebra against the gridded fair CRPS vs ERA5 at leads 0..10, both stochastic sources live; the manuscript's gridded E2E ensemble | 17 |
e2e_xanthicmarmot |
joint fine-tune of the same chain plus the tas station decoders against the station fair CRPS at leads 0..10; the manuscript's station E2E ensemble for 2 m temperature |
12 |
e2e_excellentpetrel |
same for the ws (10 m wind speed) decoders |
17 |
The _normmode parent directories contain "embed" in their name for
historical reasons; the runs inside are noise_mode="norm". The station
decoders used by the component pairs and the deterministic encoder/processor
are the upstream release (https://huggingface.co/datasets/av555/aardvark-weather).
Evaluation caches (results/)
All caches use the fair CRPS estimator and carry per-initialisation score series, verification-rank histograms and event-probability histograms (10th/90th-percentile events of the 2007-2017 climatology) for the significance tests, rank histograms and decision-value analysis.
| File | Content |
|---|---|
e2e_gridded_leads_cobaltmillipede_7x7.npz, e2e_gridded_lead0_cobaltmillipede_7x7.npz |
joint gridded-objective fine-tune, 7x7 nested ensemble vs ERA5, 2018, leads 1..10 + its encoder's lead-0 analysis: skill, spread, ANOVA variance split, entropy/MI decomposition |
e2e_gridded_leads_{powerfulzebra,malachitegoshawk}_entropy7x7.npz |
the component pairs, same protocol |
e2e_gridded_leads_powerfulzebra_rankhist7x7.npz |
the headline pair again with fresh member draws (rank histograms) |
e2e_gridded_leads_enigmaticmalkoha_entropy7x7.npz |
control processor composed with the stochastic encoder, 7x7 |
e2e_gridded_leads_{powerfulzebra,detic_enigmaticmalkoha}_detenc_1x49.npz |
dropout-only arms (1x49) of both processors on the identical deterministic analysis (compensation swap) |
e2e_gridded_leads_upstream.npz |
deterministic upstream baseline, same protocol |
e2e_gridded_lead0_{glaringsawfly,deviousiguana,upstream}.npz |
encoder analysis (lead 0) skill vs ERA5 |
e2e_nested_leads_{xanthicmarmot_7x7_tas,excellentpetrel_7x7_ws,cobaltmillipede_7x7_tas}.json (+ _events.npz) |
HadISD station verification of the joint fine-tunes, global and per region |
e2e_nested_leads_{powerfulzebra,malachitegoshawk}_7x7_{tas,ws}.json, e2e_nested_leads_upstream_{tas,ws}.json (+ _events.npz) |
station verification of the component pairs and the deterministic baseline |
e2e_station_leads_xanthicmarmot_7x7_tas.npz, e2e_station_leads_powerfulzebra_7x7_{tas,ws}.npz |
per-station metrics for the maps |
ose_smoke_full_joint_grid_7x7.json (+ _maps.npz), ose_smoke_full_{norm,embed}_7x7.json |
observation-denial (OSE) experiment, 9 streams, joint system and component pairs |
error_spectra_{cobaltmillipede_7x7,upstream}_full.npz |
zonal error spectra vs ERA5, all 2018 inits |
obs_gaps_train00z.npz |
missing-observation fractions per stream over the training years |
crossed_decomposition_powerfulzebra_crossed_fullyear.npz |
crossed (frozen dropout-mask) decomposition |
Provenance and licenses
The fine-tuned weights start from the Aardvark Weather release
(Allen et al., Nature 2025) and were trained on the Aardvark observational
dataset (CC BY-NC-ND 4.0) with ERA5 targets (Copernicus/ECMWF, via
WeatherBench 2). Please cite those sources alongside this dataset and the
manuscript (citation metadata in the code repository's CITATION.cff).
- Downloads last month
- 206