Only warn if atomic types are present in the validation dataset but n…

…ot in the training dataset

src/metatrain/utils/composition.py

@@ -1,3 +1,4 @@
+import warnings
 from typing import Dict, List, Optional, Union
 
 import torch
@@ -85,9 +86,10 @@ def train_model(
 
         missing_types = sorted(set(self.atomic_types) - set(get_atomic_types(datasets)))
         if missing_types:
-            raise ValueError(
+            warnings.warn(
                 f"Provided `datasets` do not contain atomic types {missing_types}. "
-                f"Known types from initilaization are {self.atomic_types}."
+                f"Known types from initilaization are {self.atomic_types}.",
+                stacklevel=2,
             )
 
         # Fill the weights for each target in the dataset info

tests/resources/generate-outputs.sh

@@ -7,5 +7,5 @@ ROOT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 
 cd $ROOT_DIR
 
-mtt train options.yaml -o model-32-bit.pt -r base_precision=32 > /dev/null
-mtt train options.yaml -o model-64-bit.pt -r base_precision=64 > /dev/null
+mtt train options.yaml -o model-32-bit.pt -r base_precision=32 #> /dev/null
+mtt train options.yaml -o model-64-bit.pt -r base_precision=64 #> /dev/null

tests/utils/test_composition.py

@@ -336,8 +336,8 @@ def test_composition_model_missing_types():
             ),
         ),
     )
-    with pytest.raises(
-        ValueError,
+    with pytest.warns(
+        UserWarning,
         match="do not contain atomic types",
     ):
         composition_model.train_model(dataset)