Add SLAC FEL model files and preliminary workflow

[pluflou]

Summary

This pull request introduces support for the SLAC FEL continuous-learning experiment, including configuration, data utilities, and integration with the main example harness. The changes add a new TOML config file, a data utility module, and update the configuration and dependency management to support the new experiment.

Approach

• Added examples/slac_fel/slac-fel.toml configuration file with model, data, training, continual learning, drift detection, logging, and visualization settings for the SLAC FEL experiment.
• Added examples/slac_fel/utils.py with data-loading and preprocessing utilities, including dataset, dataloader, feature config, scaler loading, window discovery, and legacy window splitting for SLAC FEL data.
• Updated examples/utils.py to support the SLAC FEL experiment by returning SLAC_FEL harness when cfg.data.name is slac-fel.

Configuration and dependency updates:

• Extended ModelCfg and DataCfg dataclasses in src/config/configuration.py to support optional model config paths and window size parameters for time-windowed datasets.
• Updated pyproject.toml to include new dependencies (types-pyyaml, botorch) required for SLAC FEL experiment and added the examples directory to the package list.

This PR also includes a fix for #85 by @Zilinghan

Dependencies

• Added: botorch
  note: can work on implementation to remove the need for it

Testing Plan

• Unit tests
• Integration tests
• e2e / smoke test
• Manual steps: python -m app --help

Documentation

• Docstrings updated
• User docs / README updated
• CHANGELOG entry

Checklist

• Code formatted (Ruff) → ruff format --check
• Lint passes (Ruff) → ruff check .
• Types pass (mypy/pyright) → mypy src
• Tests pass (pytest) → pytest -q

• Backward compatibility considered
• Adequate comments for tricky parts
• CI green

[anagainaru]

Looks good, I recommend the two changes so you pass the CI. I cannot run this since I don't have your data/model but I changed it to work with my model and it seems like it's running fine.

In addition to the changes requested, you need types-pyyaml in pyproject.

diff --git a/pyproject.toml b/pyproject.toml
index 8b252e9..e3657bf 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -20,7 +20,8 @@ dependencies = [
     "transformers (>=4.57.1,<5.0.0)",
     "river (>=0.21.0,<0.24.0)",
     "evidently (>=0.4.0,<0.8.0)",
-    "matplotlib (>=3.10.7,<4.0.0)"
+    "matplotlib (>=3.10.7,<4.0.0)",
+    "types-pyyaml (>=6.0.12.20250915,<7.0.0.0)"
 ]

 [dependency-groups]
diff --git a/poetry.lock b/poetry.lock
index 426ce3d..c232f3e 100644
--- a/poetry.lock
+++ b/poetry.lock
@@ -4874,6 +4874,18 @@ files = [
     {file = "types_pytz-2025.2.0.20251108.tar.gz", hash = "sha256:fca87917836ae843f07129567b74c1929f1870610681b4c92cb86a3df5817bdb"},
 ]

+[[package]]
+name = "types-pyyaml"
+version = "6.0.12.20250915"
+description = "Typing stubs for PyYAML"
+optional = false
+python-versions = ">=3.9"
+groups = ["main"]
+files = [
+    {file = "types_pyyaml-6.0.12.20250915-py3-none-any.whl", hash = "sha256:e7d4d9e064e89a3b3cae120b4990cd370874d2bf12fa5f46c97018dd5d3c9ab6"},
+    {file = "types_pyyaml-6.0.12.20250915.tar.gz", hash = "sha256:0f8b54a528c303f0e6f7165687dd33fafa81c807fcac23f632b63aa624ced1d3"},
+]
+
 [[package]]
 name = "types-requests"
 version = "2.32.4.20250913"

Let me know when it's ready to be merged and I will approve.

[anagainaru]

diff --git a/examples/slac_fel/utils.py b/examples/slac_fel/utils.py
index e52df9c..59d347d 100644
--- a/examples/slac_fel/utils.py
+++ b/examples/slac_fel/utils.py
@@ -181,8 +181,8 @@ def load_fel_data(
     X_raw = torch.tensor(df[input_cols].values, dtype=torch.float32)
     y_raw = torch.tensor(df[output_cols].values, dtype=torch.float32)

-    X_scaled: Tensor = input_scaler.transform(X_raw)
-    y_scaled: Tensor = output_scaler.transform(y_raw)
+    X_scaled: Tensor = input_scaler(X_raw)
+    y_scaled: Tensor = output_scaler(y_raw)

     return X_scaled, y_scaled, df.index

[pluflou]

these are botorch transformers right now (from the model source). I can save and load them differently if we don't want the added dependency

[anagainaru]

The CI was complaining, it's prob not clear to mypy what those are.

[pluflou]

I updated the way the model is being loaded (there was a bug before), and the config for the drift detector (not confident that these make sense though). There seems to be something off with the logging, so I don't want to mark it as ready to merge yet.

[anagainaru]

I updated the way the model is being loaded (there was a bug before), and the config for the drift detector (not confident that these make sense though). There seems to be something off with the logging, so I don't want to mark it as ready to merge yet.

I will try to run it as well tomorrow before our meeting

[pluflou]

I updated the docstrings to be in Google style, which seems to be the dominant/intended format across the codebase (although config/configuration.py and some example modules are using NumPy style)

[anagainaru]

After I rebase this on main you can use in the toml file the following option (introduced in #88):

max_ckpts = 1
ckpts_path = "output/slac/"

to save checkpoints after each CL loop and only keep the most recent one.