Add backend routing, transform caching, and sklearn estimator caching

.gitignore

@@ -2,11 +2,15 @@
 .cache/
 backbone_cache/
 backend_store/
+sklearn_cache/
+ituna_cache/
 checkpoints/
 data/
+data
 logs/
 optuna_studies/
 slurm_output/
+slurm_output
 wandb/*
 !wandb/*.py
 !wandb/*.sh
@@ -204,6 +208,7 @@ cython_debug/
 #  and can be added to the global gitignore or merged into this file. However, if you prefer,
 #  you could uncomment the following to ignore the entire vscode folder
 .vscode/
+.cursor/
 
 # Ruff stuff:
 .ruff_cache/

AUTHORS.md

@@ -2,8 +2,8 @@
 
 ## v0.1.0 Development
 
-🐟<b>i</b><span style="color: #3C92ED;">Tuna</span> was initially developed and is being maintained by Tobias Schmidt ([&#64;TobiasSchmidtDE](https://github.com/TobiasSchmidtDE)) and Steffen Schneider ([&#64;stes](https://github.com/stes)) in the [Dynamical Inference Lab](https://dynamical-inference.ai/) at Helmholtz Munich. 
+🐟<b>i</b><span style="color: #3C92ED;">Tuna</span> was initially developed and is being maintained by Tobias Schmidt ([&#64;TobiasSchmidtDE](https://github.com/TobiasSchmidtDE)) and Steffen Schneider ([&#64;stes](https://github.com/stes)) in the [Dynamical Inference Lab](https://dynamical-inference.ai/) at Helmholtz Munich.
 
 Lilly May ([&#64;Lilly-May](https://github.com/Lilly-May)) contributed consistency metrics for sparse autoencoders (to be released soon).
 
-Paul Pommer ([&#64;ppommer](https://github.com/ppommer)) contributed demos for [piVAE](https://github.com/zhd96/pi-vae) and [FastSAE](https://github.com/dynamical-inference/fastsae) (to be released soon) and alpha-tested the package.
+Paul Pommer ([&#64;ppommer](https://github.com/ppommer)) contributed demos for [piVAE](https://github.com/zhd96/pi-vae) and [FastSAE](https://github.com/dynamical-inference/fastsae) (to be released soon) and alpha-tested the package.

CONTRIBUTING.md

@@ -119,8 +119,8 @@ The documentation is built using [Jupyter Book](https://jupyterbook.org/) versio
 ### Local Build
 
 ```bash
-# Install jupyter-book (must be version <2)
-pip install "jupyter-book<2"
+# Install documentation dependencies (includes jupyter-book<2 and jupytext)
+pip install -e ".[docs]"
 
 # Build the docs from the project root
 jupyter-book build .
@@ -131,6 +131,29 @@ open _build/html/index.html   # macOS
 xdg-open _build/html/index.html  # Linux
 ```
 
+### Editing Notebooks (Jupytext, Recommended)
+
+We keep tutorial/demo notebooks paired as:
+
+- `*.ipynb` (rendered by Jupyter Book)
+- `*.py` in **percent** format (easy to diff/review)
+
+To edit a notebook:
+
+1. Edit the corresponding `*.py` percent file.
+2. Sync back to the `*.ipynb`:
+
+```bash
+jupytext --sync docs/tutorials/<notebook>.ipynb
+```
+
+To pair a new notebook (creates/updates the `*.py` alongside the `*.ipynb`):
+
+```bash
+jupytext --set-formats ipynb,py:percent docs/tutorials/new_notebook.ipynb
+jupytext --sync docs/tutorials/new_notebook.ipynb
+```
+
 ### Local Server
 
 To serve the docs locally with live preview:
@@ -161,7 +184,7 @@ Press Ctrl+C to stop.
 
 ```bash
 # Install required tools
-pip install "jupyter-book<2" ghp-import
+pip install -e ".[docs]"
 
 # Build and publish
 jupyter-book build .

README.md

@@ -91,6 +91,7 @@ Full documentation is available at **[dynamical-inference.github.io/ituna](https
 - **Quickstart notebook**: [`docs/tutorials/quickstart.ipynb`](docs/tutorials/quickstart.ipynb) - minimal working example
 - **Core concepts**: [`docs/tutorials/core.ipynb`](docs/tutorials/core.ipynb) - in-depth walkthrough
 - **Backends**: [`docs/tutorials/backends.ipynb`](docs/tutorials/backends.ipynb) - caching and distributed execution
+- **sklearn caching**: [`docs/guides/sklearn_caching.ipynb`](docs/guides/sklearn_caching.ipynb) - cache standalone sklearn estimators (`fit`/`transform`/`predict`/`score`)
 
 ## Backends
 
@@ -118,6 +119,10 @@ with config.config_context(
     BACKEND_KWARGS={"trigger_type": "auto", "num_workers": 4},
 ):
     ensemble.fit(X)
+
+# Advanced: route different operations to different backends
+# (e.g. distributed estimator fit + locally cached consistency transforms)
+# config.register_backend_route(method="fit", model_class=metrics.ConsistencyTransform, backend="disk_cache")
 ```
 
 ### CLI Commands

_config.yml

@@ -11,6 +11,7 @@ repository:
 
 exclude_patterns:
   - .github/*
+  - .cursor/**
   - slurm/*
   - third_party/*
   - _build/*

_toc.yml

@@ -6,6 +6,9 @@ parts:
       - file: docs/tutorials/quickstart
       - file: docs/tutorials/core
       - file: docs/tutorials/backends
+  - caption: Guides
+    chapters:
+      - file: docs/guides/sklearn_caching
   - caption: Demos
     chapters:
       - file: docs/demos/cebra_best_practices