🚀 Make ProblemDefinition a Pydantic model

docs/source/core_concepts/disk_format.md

@@ -18,8 +18,7 @@ folder
 │   │   └── sample_yyyyyyyyy
 │   └── infos.yaml
 └── problem_definition
-    ├── problem_infos.yaml
-    └── split.json (or split.csv for <=0.1.7)
+    └── problem_infos.yaml
 ```
 
 - `dataset/samples/`: one directory per {py:class}`~plaid.containers.sample.Sample`.

docs/source/core_concepts/feature_identifiers.md

@@ -121,7 +121,7 @@ Legacy name-based methods (e.g., `add_input_scalars_names`) are deprecated; pref
 - Always include enough context to disambiguate a feature. For fields/nodes on multiple bases/zones/times, set all relevant keys.
 - Use {py:meth}`~plaid.containers.sample.Sample.get_all_features_identifiers()` to introspect what identifiers exist in a sample.
 - Use sets to deduplicate identifiers safely: `set(list_of_identifiers)`.
-- When authoring problem definitions on disk, {py:meth}`~plaid.problem_definition.ProblemDefinition._save_to_dir_` persists identifiers under `problem_definition/problem_infos.yaml` (keys `input_features` and `output_features`).
+- When authoring problem definitions on disk, {py:meth}`~plaid.problem_definition.ProblemDefinition.save_to_dir` persists identifiers under `problem_definition/problem_infos.yaml` (keys `input_features` and `output_features`).
 
 ## See also
 

docs/source/core_concepts/problem_definition.md

@@ -27,7 +27,10 @@ pb.add_out_feature_identifier(FeatureIdentifier({
 splits = {"train": [0, 1, 2], "test": [3, 4]}
 pb.set_split(splits)
 
-pb._save_to_dir_("problem_definition")
+pb.save_to_dir("problem_definition")
+
+# later
+pb2 = ProblemDefinition.load("problem_definition")
 ```
 
 {py:class}`~plaid.problem_definition.ProblemDefinition` supports filtering helpers to intersect existing inputs/outputs with a candidate list of identifiers.

examples/post/bisect_example.py

@@ -47,7 +47,7 @@
 # Load PLAID datasets and problem metadata objects
 ref_ds = Dataset(dataset_directory / "dataset_ref")
 pred_ds = Dataset(dataset_directory / "dataset_near_pred")
-problem = ProblemDefinition(dataset_directory / "problem_definition")
+problem = ProblemDefinition.load(dataset_directory / "problem_definition")
 
 # Get output scalars from reference and prediction dataset
 ref_out_scalars, pred_out_scalars, out_scalars_names = prepare_datasets(
@@ -98,7 +98,7 @@
 # Load PLAID datasets and problem metadata objects
 ref_path = Dataset(dataset_directory / "dataset_ref")
 pred_path = Dataset(dataset_directory / "dataset_pred")
-problem_path = ProblemDefinition(dataset_directory / "problem_definition")
+problem_path = ProblemDefinition.load(dataset_directory / "problem_definition")
 
 # Using PLAID objects to generate bisect plot on feature_2
 plot_bisect(ref_path, pred_path, problem_path, "feature_2", "equal_bisect_plot")
@@ -114,7 +114,7 @@
 # Mix
 ref_path = dataset_directory / "dataset_ref"
 pred_path = dataset_directory / "dataset_near_pred"
-problem_path = ProblemDefinition(dataset_directory / "problem_definition")
+problem_path = ProblemDefinition.load(dataset_directory / "problem_definition")
 
 # Using scalar index and verbose option to generate bisect plot
 scalar_index = 0
@@ -129,4 +129,4 @@
 
 os.remove("converge_bisect_plot.png")
 os.remove("differ_bisect_plot.png")
-os.remove("equal_bisect_plot.png")
+os.remove("equal_bisect_plot.png")

examples/post/metrics_example.py

@@ -46,7 +46,7 @@
 # Load PLAID datasets and problem metadata objects
 ref_ds = Dataset(dataset_directory / "dataset_ref")
 pred_ds = Dataset(dataset_directory / "dataset_near_pred")
-problem = ProblemDefinition(dataset_directory / "problem_definition")
+problem = ProblemDefinition.load(dataset_directory / "problem_definition")
 
 # Get output scalars from reference and prediction dataset
 ref_out_scalars, pred_out_scalars, out_scalars_names = prepare_datasets(
@@ -102,7 +102,7 @@
 # Load PLAID datasets and problem metadata objects
 ref_ds = Dataset(dataset_directory / "dataset_ref")
 pred_ds = Dataset(dataset_directory / "dataset_pred")
-problem = ProblemDefinition(dataset_directory / "problem_definition")
+problem = ProblemDefinition.load(dataset_directory / "problem_definition")
 
 # Pretty print activated with verbose mode
 metrics = compute_metrics(ref_ds, pred_ds, problem, "second_metrics", verbose=True)
@@ -123,4 +123,4 @@
 pretty_metrics(dictionary)
 
 os.remove("first_metrics.yaml")
-os.remove("second_metrics.yaml")
+os.remove("second_metrics.yaml")

examples/problem_definition_example.py

@@ -154,7 +154,7 @@
 # ### Load a ProblemDefinition from a directory via initialization
 
 # %%
-problem = ProblemDefinition(pb_def_save_fname)
+problem = ProblemDefinition.load(pb_def_save_fname)
 print(problem)
 
 # %% [markdown]
@@ -168,6 +168,5 @@
 # ### Load from a directory via a Dataset instance
 
 # %%
-problem = ProblemDefinition()
-problem.load(pb_def_save_fname)
+problem = ProblemDefinition.load(pb_def_save_fname)
 print(problem)

src/plaid/bridges/huggingface_bridge.py

@@ -375,8 +375,7 @@ def load_problem_definition_from_hub(
     with open(yaml_path, "r", encoding="utf-8") as f:
         yaml_data = yaml.safe_load(f)
 
-    prob_def = ProblemDefinition()
-    prob_def._initialize_from_problem_infos_dict(yaml_data)
+    prob_def = ProblemDefinition.model_validate(yaml_data)
 
     return prob_def
 
@@ -484,9 +483,7 @@ def load_problem_definition_from_disk(
     Returns:
         ProblemDefinition: The loaded problem definition.
     """
-    pb_def = ProblemDefinition()
-    pb_def._load_from_file_(Path(path) / Path("problem_definitions") / Path(name))
-    return pb_def
+    return ProblemDefinition.load(Path(path) / Path("problem_definitions") / Path(name))
 
 
 def load_tree_struct_from_disk(
@@ -698,19 +695,33 @@ def huggingface_description_to_problem_definition(
     problem_definition = ProblemDefinition()
     for func, key in [
         (problem_definition.set_task, "task"),
+        (problem_definition.set_score_function, "score_function"),
         (problem_definition.set_split, "split"),
+    ]:
+        if key in description:
+            func(description[key])
+
+    if "input_features" in description:
+        problem_definition.add_in_features_identifiers(description["input_features"])
+    if "output_features" in description:
+        problem_definition.add_out_features_identifiers(description["output_features"])
+    if "constant_features" in description:
+        problem_definition.add_constant_features_identifiers(
+            description["constant_features"]
+        )
+    legacy_keys = [
         (problem_definition.add_input_scalars_names, "in_scalars_names"),
         (problem_definition.add_output_scalars_names, "out_scalars_names"),
         (problem_definition.add_input_fields_names, "in_fields_names"),
         (problem_definition.add_output_fields_names, "out_fields_names"),
+        (problem_definition.add_input_timeseries_names, "in_timeseries_names"),
+        (problem_definition.add_output_timeseries_names, "out_timeseries_names"),
         (problem_definition.add_input_meshes_names, "in_meshes_names"),
         (problem_definition.add_output_meshes_names, "out_meshes_names"),
-    ]:
-        try:
+    ]
+    for func, key in legacy_keys:
+        if key in description:
             func(description[key])
-        except KeyError:
-            logger.error(f"Could not retrieve key:'{key}' from description")
-            pass
 
     return problem_definition
 

src/plaid/post/bisect.py

@@ -104,7 +104,7 @@ def plot_bisect(
     if isinstance(pred_dataset, (str, Path)):
         pred_dataset: Dataset = Dataset(pred_dataset)
     if isinstance(problem_def, (str, Path)):
-        problem_def: ProblemDefinition = ProblemDefinition(problem_def)
+        problem_def: ProblemDefinition = ProblemDefinition.load(problem_def)
 
     # Load the testing_set
     # testing_set = problem_def.get_split("test")

src/plaid/post/metrics.py

@@ -148,7 +148,7 @@ def compute_metrics(
     if isinstance(pred_dataset, (str, Path)):
         pred_dataset: Dataset = Dataset(pred_dataset)
     if isinstance(problem, (str, Path)):
-        problem: ProblemDefinition = ProblemDefinition(problem)
+        problem: ProblemDefinition = ProblemDefinition.load(problem)
 
     ### Get important formated values ###
     problem_split = problem.get_split()