Fix composition weights schema

docs/src/architectures/soap-bpnn.rst

@@ -147,6 +147,7 @@ The parameters for training are
 :param learning_rate: learning rate
 :param log_interval: number of epochs that elapse between reporting new training results
 :param checkpoint_interval: Interval to save a checkpoint to disk.
+:param fixed_composition_weights: TODO including an example.
 :param per_atom_targets: Specifies whether the model should be trained on a per-atom
     loss. In that case, the logger will also output per-atom metrics for that target. In
     any case, the final summary will be per-structure.

src/metatrain/experimental/soap_bpnn/schema-hypers.json

@@ -120,8 +120,14 @@
         "fixed_composition_weights": {
           "type": "object",
           "patternProperties": {
-            "^[0-9]+$": {
-              "type": "number"
+            "^.*$": {
+              "type": "object",
+              "patternProperties": {
+                "^[0-9]+$": {
+                  "type": "number"
+                }
+              },
+              "additionalProperties": false
             }
           },
           "additionalProperties": false

src/metatrain/experimental/soap_bpnn/tests/test_functionality.py

@@ -195,3 +195,7 @@ def test_output_per_atom():
 
     assert outputs["energy"].block().samples.names == ["system", "atom"]
     assert outputs["energy"].block().values.shape == (4, 1)
+
+
+def test_fixed_composition_weights():
+    pass