Implementation of NormalizedReluBounding for non-zero thresholds

[lzampier]

(pull request already under review in the anemoi-models repository)

This new approach can be used when applying the Relu bounding to a threshold value other than the default 0. An obvious use case is the ocean temperature, which should not go below the freezing temperature of the ocean (approx. 271.15 K). The threshold imputed to the function is normalised according to the normalisation scheme specified in the config file.

Here is an example call to be added to the config:

    - _target_: anemoi.models.layers.bounding.NormalizedReluBounding
      variables:
      - avg_tos
      min_val: [271.15] # Needs to be a list 
      normalizer: ["mean-std"]

In case of multiple variables

    - _target_: anemoi.models.layers.bounding.NormalizedReluBounding
      variables:
      - var1
      - var2
      min_val: [val1, val2] # Needs to be a list 
      normalizer: ["norm1", "norm2"]

[FussyDuck]

All committers have signed the CLA.

[jakob-schloer]

Great work!

Could you add an example in the config? E.g.

    - _target_: anemoi.models.layers.bounding.NormalizedReluBounding
      variables: [sst]
      min_val: [-2]
      normalizer: ['mean-std']

Could be also commented out.

[lzampier]

Great work!

Could you add an example in the config? E.g.

    - _target_: anemoi.models.layers.bounding.NormalizedReluBounding
      variables: [sst]
      min_val: [-2]
      normalizer: ['mean-std']

Could be also commented out.

Thank you, @jakob-schloer. I have added the commented example in the config.

[jakob-schloer]

Very nice work! Thanks for including my suggestions.