Standardise configs/models for comparison#241

Merged

jemrobinson merged 32 commits intomainfrom

model-comparison

Mar 25, 2026

Commits on Mar 5, 2026

Commits on Mar 6, 2026

:label: Add type hints for tensor shapes to DDPM
jemrobinson
committed
:coffin: Remove slurm specific options from DDPM model
jemrobinson
committed
:wrench: Drop convolutional size in piecewise encoders to reduce memory usage
jemrobinson
committed

Commits on Mar 9, 2026

:wrench: Add a config for predicting 21 days into the future
jemrobinson
committed
:rewind: Hemisphere is no longer needed as a grouping
jemrobinson
committed
:wrench: Reduce size of CNN latent space to support 21 forecast days
jemrobinson
committed
:wrench: Reduce number of channels in naive-unet-naive to support 21 day forecasts
jemrobinson
committed
:wrench: Fix config name
jemrobinson
committed
:wrench: Use 2 convolutional blocks in piecewise encoder by default
jemrobinson
committed
:wrench: Use 3 convolutional blocks and clamp in piecewise decoder by default
jemrobinson
committed
:rotating_light: Fix linting/testing errors
jemrobinson
committed

Commits on Mar 20, 2026

:wrench: updating predict config for the demo
IFenton
committed

Commits on Mar 23, 2026

:wrench: Add a new model called 'quick_test' that runs faster at the cost of accuracy
jemrobinson
committed

Commits on Mar 25, 2026

:memo: Added note about default quick test model
jemrobinson
committed