Skip to content

[WIP] Standard baseline configs#324

Draft
jemrobinson wants to merge 23 commits into
mainfrom
add-baseline-configs
Draft

[WIP] Standard baseline configs#324
jemrobinson wants to merge 23 commits into
mainfrom
add-baseline-configs

Conversation

@jemrobinson

@jemrobinson jemrobinson commented Jun 25, 2026

Copy link
Copy Markdown
Member

This PR adds some baseline configs for the model comparisons discussed in #323.

TL;DR

  • nothing is beating persistence on a two day forecast (perhaps we want to look at e.g. 7 days?)
  • some very weird aliasing features on piecewise and vit that make me feel that we could better-optimise these defaults
  • is combining train and validation on the metric plots actually useful or is it unnecessary noise?

Models

00: naive-unet-naive

north south

01: persistence

north south

02: cnn-unet-cnn

north south

03: cnn-vit-cnn

north south

04: ddpm

north south

05: piecewise-unet-piecewise-5336004.out

north south

Metrics

RMSE

north-rmse south-rmse

SIE Error

north-sieerror south-sieerror

MAE

north-mae south-mae

@jemrobinson jemrobinson changed the title Standard baseline configs [WIP] Standard baseline configs Jun 25, 2026
@jemrobinson jemrobinson marked this pull request as draft June 25, 2026 13:39
@IFenton

IFenton commented Jun 26, 2026

Copy link
Copy Markdown
Contributor

Couple of quick thoughts.

nothing is beating persistence on a two day forecast (perhaps we want to look at e.g. 7 days?)

Looking at 7 days seems sensible anyway, as you can then get an idea of how the forecast accuracy develops over time. (And as you say it is more likely to beat persistence).

As we've discussed a bit elsewhere (#323), doing a bit of work to optimise the models before we do this seems sensible. E.g. Erin's trained ViT model (https://wandb.ai/turing-seaice/evaluate/runs/0r9jomqk) does beat persistence

@jemrobinson

Copy link
Copy Markdown
Member Author

For reference, @erinuclkwon's model was using

n_forecast_steps: 7
n_history_steps: 3

which is what the next round of testing will use.

@jemrobinson jemrobinson force-pushed the add-baseline-configs branch from 57f3cfb to b545ae1 Compare June 26, 2026 15:26
@github-actions

Copy link
Copy Markdown

Coverage report

Click to see where and how coverage changed

FileStatementsMissingCoverageCoverage
(new stmts)
Lines missing
  icenet_mp
  model_service.py 31-33, 43-45, 213-221
  icenet_mp/callbacks
  ema_weight_averaging_callback.py 29-30, 36-37, 41-42
  metric_summary_callback.py 29, 62
  icenet_mp/compatibility
  __init__.py
  icenet_mp/data_processors
  data_downloader.py 190-197, 202-212, 231-235, 245-249
  icenet_mp/models
  base_model.py 95-96
  persistence.py
  icenet_mp/models/common
  conv_block_downsample.py
  conv_block_upsample.py
  icenet_mp/models/decoders
  cnn_decoder.py
  icenet_mp/models/encoders
  cnn_encoder.py
  icenet_mp/models/processors
  vit.py 41-42
  icenet_mp/types
  __init__.py
  simple_datatypes.py
Project Total  

This report was generated by python-coverage-comment-action

@jemrobinson jemrobinson force-pushed the add-baseline-configs branch from 421c0b8 to 4dbf8d7 Compare June 29, 2026 15:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants