refactor: Improve config parallelization validation with detailed err…

…or reporting

saev/training.py

@@ -414,14 +414,29 @@ def __iter__(self):
 def split_cfgs(cfgs: list[config.Train]) -> list[list[config.Train]]:
     """
     Splits configs into groups that can be parallelized.
+
+    Arguments:
+        A list of configs from a sweep file.
+
+    Returns:
+        A list of lists, where the configs in each sublist do not differ in any keys that are in `CANNOT_PARALLELIZE`. This means that each sublist is a valid "parallel" set of configs for `train`.
     """
-    # Group configs by n_workers value
-    groups = collections.defaultdict(list)
+
+    seen = collections.defaultdict(list)
     for cfg in cfgs:
-        groups[cfg.n_workers].append(cfg)
-
-    # Return list of groups
-    return list(groups.values())
+        dct = dataclasses.asdict(cfg)
+        dct = helpers.flattened(dct)
+        for key, value in dct.items():
+            seen[key].append(value)
+
+    bad_keys = {}
+    for key, values in seen.items():
+        if key in CANNOT_PARALLELIZE and len(set(values)) != 1:
+            bad_keys[key] = values
+
+    if bad_keys:
+        msg = ", ".join(f"'{key}': {values}" for key, values in bad_keys.items())
+        raise ValueError(f"Cannot parallelize training over: {msg}")
 
 
 ##############