Skip to content

make srdrwfy6 inference work - #2674

Open
Jubeku wants to merge 2 commits into
develop-ssl-diffusion-v1from
jk/develop-ssl-diffusion-v1/inf-srdrwfy6
Open

make srdrwfy6 inference work#2674
Jubeku wants to merge 2 commits into
develop-ssl-diffusion-v1from
jk/develop-ssl-diffusion-v1/inf-srdrwfy6

Conversation

@Jubeku

@Jubeku Jubeku commented Jul 22, 2026

Copy link
Copy Markdown
Contributor

Inference is run as follows:

uv run inference --from-run-id srdrwfy6 --options test_config.samples_per_mini_epoch=16 test_config.output.num_samples=16 test_config.start_date=2023-10-01T00:00 test_config.end_date=2023-12-31T00:00 test_config.time_window_step=06:00:00

Results into:

validation (aig5t8q3) : 000 : 
                        0.07851782130698363
LossPhysical.ERA5.mse.avg : 7.8518E-02 
LossPhysical.loss_avg : 7.8518E-02 


100%|████████████████████████████████████████████████████████████████████████| 12/12 [05:18<00:00, 26.53s/it]
Finished inference run with id: aig5t8q3

Seb's baseline:

100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 16/16 [07:50<00:00, 26.21s/it]validation (hw12itcf) : 000 : 
                        0.05010054318699986
LossPhysical.ERA5.mse.avg : 5.0101E-02 
LossPhysical.loss_avg : 5.0101E-02 


100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 16/16 [07:50<00:00, 29.42s/it]
Finished inference run with id: hw12itcf

Description

Issue Number

Is this PR a draft? Mark it as draft.

Checklist before asking for review

  • I have performed a self-review of my code
  • My changes comply with basic sanity checks:
    • I have fixed formatting issues with ./scripts/actions.sh lint
    • I have run unit tests with ./scripts/actions.sh unit-test
    • I have documented my code and I have updated the docstrings.
    • I have added unit tests, if relevant
  • I have tried my changes with data and code:
    • I have run the integration tests with ./scripts/actions.sh integration-test
    • (bigger changes) I have run a full training and I have written in the comment the run_id(s): launch-slurm.py --time 60
    • (bigger changes and experiments) I have shared a hegdedoc in the github issue with all the configurations and runs for this experiments
  • I have informed and aligned with people impacted by my change:
    • for config changes: the MatterMost channels and/or a design doc
    • for changes of dependencies: the MatterMost software development channel

@clessig clessig left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks. Minor comments

Comment thread src/weathergen/train/trainer.py
# not present here and are skipped instead of raising a ValueError.
requested = stream_info.get("geoinfo_channels")
missing = [ch for ch in requested if ch not in ds.variables]
if missing:

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if len(missing) > 0 :

Comment thread src/weathergen/utils/validation_io.py Outdated
n_pred_steps = len(model_output.physical)
if n_pred_steps > len(timestep_idxs):
timestep_idxs = list(range(forecast_offset, forecast_offset + n_pred_steps))
# # Diffusion inference inflates the model output's fstep dimension to one entry per

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it useful to keep this for debug purposes?

#2676)

* draft of changes to make inference with develop-ssl trained srdrwfy6 work

* Restore _expand_targets_to_match_preds function

Restore target expansion logic to handle multiple forecast steps and improve clarity in the warning messages.

---------

Co-authored-by: Sebastian Hickman <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: No status

Development

Successfully merging this pull request may close these issues.

2 participants