Merge pull request #225 from automl/216-extend-converter-documentation-for-more-detailed-capabilities-and-limitations

added documentation to converter

Author: mwever

CHANGELOG.md

@@ -1,8 +1,15 @@
+# Version 1.3.8
+
+## Converter Documentation
+- Added documentation to make it visible which converters cant handle instances
+- Added documentation to give some information about limitations (and capabilities) of the converters
+
 # Version 1.3.7
 
 ## Documentation
 - Added the hint that for the api example the example records need to be run first to create the run.
 
+
 # Version 1.3.6
 
 ## New SMAC version compatibility

docs/converters/amltk.rst

@@ -35,4 +35,7 @@ consider when running AMLTK:
 
    - Alternatively, you can still manually add the DeepCave time columns to ``history.parquet``
      after the AMLTK run has finished by loading it into a Pandas Dataframe, manipulating it, and
-     writing it back to the ``history.parquet`` file.
+     writing it back to the ``history.parquet`` file.
+
+.. note::
+   Instances are not supported.

docs/converters/dataframe.rst

@@ -48,5 +48,8 @@ SUCCESS, TIMEOUT, MEMORYOUT, CRASHED, ABORTED, NOT_EVALUATED, FAILED, PRUNED, UN
 The ``budget`` column should contain the multi-fidelity budget and can be omitted if not used.
 The same holds for the ``seed`` column, containing the trial seed.
 
-.. warning::
+.. note::
     Conditions and forbiddens are not supported in the current version of the Pandas DataFrame converter.
+    Weights are not supported by us. They will be ignored.
+    hyperparameters are expected to be of type float, categorical or integer.
+    The numeric distribution must be normal, beta or uniform.

docs/converters/optuna.rst

@@ -31,8 +31,11 @@ Furthermore, to load an Optuna run into DeepCAVE, it is necessary to install Opt
 
         pip install deepcave[optuna]
 
-.. warning::
-    Loading Optuna runs with conditional search spaces or dynamic hyperparameter value ranges
-    is not supported.
-    This limitation arises because Optuna leverages a dynamic search space, which cannot be
-    trivially converted into a static search space as used in DeepCAVE.
+.. note::
+   Optuna does not support instances.
+   The hyperparameters in the Optuna study must be of type FloatDistribution, IntDistribution or CategoricalDistribution.
+   Loading Optuna runs with conditional search spaces or dynamic hyperparameter value ranges
+   is not supported.
+   This limitation arises because Optuna leverages a dynamic search space, which cannot be
+   trivially converted into a static search space as used in DeepCAVE.
+

docs/converters/smac.rst

@@ -15,3 +15,5 @@ as well as the run history:
       "configspace.json"
       "runhistory.json"
 
+.. note::
+   Instances are not supported.