[minor] fix for GLM4.7 mtp module in PTQ

examples/llm_ptq/example_utils.py

@@ -312,13 +312,13 @@ def get_inlined_mtp_prefixes(config: Any) -> list[str]:
 def _keys_to_prefixes(keys: Iterable[str]) -> set[str]:
     """Invert separate-file MTP keys into the prefixes the exporter needs for exclude_modules.
     ``"mtp.fc.weight"`` → ``{"mtp"}``; ``"mtp.layers.0.q_proj.weight"`` →
-    ``{"mtp", "mtp.layers.0"}``. Caller must filter out inlined keys; otherwise
-    ``"model.layers.78.eh_proj.weight"`` would emit ``"model"`` as a prefix.
+    ``{"mtp", "mtp.layers.0"}``. ``"model"`` top-level is dropped to avoid the
+    ``"model*"`` wildcard covering the whole backbone.
     """
     prefixes: set[str] = set()
     for key in keys:
         parts = key.split(".")
-        if parts:
+        if parts and parts[0] != "model":
             prefixes.add(parts[0])
         for i, part in enumerate(parts):
             if part == "layers" and i + 1 < len(parts) and parts[i + 1].isdigit():

tests/examples/llm_ptq/test_example_utils.py

@@ -134,6 +134,16 @@ def test_load_mtp_weights_separate_indexed_shard(tmp_path):
     assert set(orphans) == set(mtp_tensors)
 
 
+def test_keys_to_prefixes_drops_model_top_level():
+    # nvbug 6108133: inlined keys like "model.layers.92.X" must NOT emit "model"
+    # as a top-level prefix (would become "model*" excluding the whole backbone).
+    out = example_utils._keys_to_prefixes(
+        ["model.layers.92.eh_proj.weight", "mtp.fc.weight", "mtp.layers.0.q_proj.weight"]
+    )
+    assert "model" not in out
+    assert out == {"mtp", "mtp.layers.0", "model.layers.92"}
+
+
 def test_load_mtp_weights_no_mtp_returns_empty(tmp_path):
     # Also pins the ``num_nextn_predict_layers=None`` regression: some configs
     # set the field explicitly to None, which must not crash ``int(None)``.