fix: count_input_cfg_levels now resolves string file references

When input_cfg is overridden via CLI to a YAML file path (e.g.
model.train_ds.input_cfg=train_all.yaml), the level counter only saw
the top-level string and reported 1 level, causing reweight_temperature
to broadcast incorrectly. Now it loads referenced YAML files to discover
nested input_cfg keys, matching the runtime behavior of
parse_and_combine_datasets.

Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com>

Author: XuesongYang

nemo/collections/common/data/lhotse/cutset.py

@@ -476,6 +476,12 @@ def count_input_cfg_levels(config: Union[DictConfig, dict]) -> int:
     the same temperature (due to propagate_attrs.copy()), we count max depth,
     not total occurrences.
 
+    String/Path values for ``input_cfg`` are treated as file references (mirroring
+    :func:`parse_and_combine_datasets`) and loaded so that nested ``input_cfg``
+    keys inside those files are counted.  If a file cannot be loaded (e.g. the
+    path contains unresolved OmegaConf interpolations), it is conservatively
+    counted as one additional level.
+
     Args:
         config: Configuration dictionary that may contain nested 'input_cfg' keys.
 
@@ -493,13 +499,22 @@ def count_input_cfg_levels(config: Union[DictConfig, dict]) -> int:
         2
     """
 
+    def _resolve_if_path(val):
+        """If *val* is a string/Path, try to load the YAML it points to."""
+        if isinstance(val, (str, Path)):
+            try:
+                return load_yaml(str(val))
+            except Exception:
+                return val
+        return val
+
     def _max_depth(obj) -> int:
         if isinstance(obj, (dict, DictConfig)):
             depths = []
             for key, val in obj.items():
                 if key == "input_cfg":
-                    # Found input_cfg: this level counts as 1 + max depth of children
-                    depths.append(1 + _max_depth(val))
+                    resolved = _resolve_if_path(val)
+                    depths.append(1 + _max_depth(resolved))
                 else:
                     depths.append(_max_depth(val))
             return max(depths, default=0)

tests/collections/common/test_lhotse_temperature_reweighting.py

@@ -400,6 +400,40 @@ def test_with_omegaconf_dictconfig(self):
         )
         assert count_input_cfg_levels(config) == 2
 
+    def test_string_input_cfg_resolves_yaml_file(self, tmp_path):
+        """String input_cfg pointing to a YAML file is loaded and traversed."""
+        inner_yaml = tmp_path / "inner.yaml"
+        inner_yaml.write_text(
+            "- type: lhotse_shar\n  shar_path: /path1\n  weight: 1.0\n"
+            "- type: lhotse_shar\n  shar_path: /path2\n  weight: 2.0\n"
+        )
+        config = {"input_cfg": str(inner_yaml)}
+        assert count_input_cfg_levels(config) == 1
+
+    def test_two_level_string_input_cfg(self, tmp_path):
+        """Top-level string ref → YAML with groups whose input_cfg are strings."""
+        lang_yaml = tmp_path / "train_lang.yaml"
+        lang_yaml.write_text("- type: lhotse_shar\n  shar_path: /audio/cuts\n  weight: 3.5\n")
+        top_yaml = tmp_path / "train_all.yaml"
+        top_yaml.write_text(
+            f"- type: group\n  weight: 100\n  input_cfg: {lang_yaml}\n"
+            f"- type: group\n  weight: 200\n  input_cfg: {lang_yaml}\n"
+        )
+        config = {"input_cfg": str(top_yaml)}
+        assert count_input_cfg_levels(config) == 2
+
+    def test_unresolvable_string_input_cfg_counts_as_one(self):
+        """Unresolvable path (e.g. OmegaConf interpolation) counts as 1 level."""
+        config = {
+            "input_cfg": [
+                {
+                    "type": "group",
+                    "input_cfg": "${oc.env:MANIFEST_ROOT}/train_ar-AE.yaml",
+                },
+            ]
+        }
+        assert count_input_cfg_levels(config) == 2
+
 
 class TestReweightTemperatureValidation:
     def test_scalar_temperature_broadcasts_to_all_levels(self):