test(recipe): give the offline Nemotron 3 Ultra provider its layer pattern

The 384-GPU H100 SFT recipe splits Nemotron 3 Ultra's hybrid_layer_pattern
into pipeline stages, so the shared offline provider used by
test_all_recipe_factories.py needs that model's pattern, as it already has
for GLM-5.x and DeepSeek-V4. Define the 108-layer pattern once in
recipe_test_utils.py and reuse it in the two Ultra recipe test files.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Kamran Jafari <kjafarisadeg@nvidia.com>

Author: kamran-nvidia

tests/unit_tests/recipes/nemotronh/test_nemotron_3_ultra.py

@@ -36,12 +36,9 @@
     nemotron_3_ultra_pretrain_config,
     nemotron_3_ultra_sft_openmathinstruct2_packed_config,
 )
-from tests.unit_tests.recipes.recipe_test_utils import patch_recipe_module_global
-
-
-# Nemotron 3 Ultra's 108-layer decoder pattern from the Hugging Face config.
-ULTRA_HYBRID_LAYER_PATTERN = (
-    "MEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEME"
+from tests.unit_tests.recipes.recipe_test_utils import (
+    NEMOTRON_3_ULTRA_HYBRID_LAYER_PATTERN,
+    patch_recipe_module_global,
 )
 
 
@@ -50,7 +47,7 @@ class _FakeUltraProvider:
 
     def __init__(self) -> None:
         self.vocab_size = 256
-        self.hybrid_layer_pattern = ULTRA_HYBRID_LAYER_PATTERN
+        self.hybrid_layer_pattern = NEMOTRON_3_ULTRA_HYBRID_LAYER_PATTERN
 
     def finalize(self) -> None:
         return None
@@ -254,7 +251,7 @@ def test_openmath_sft_uses_48_node_layout() -> None:
     assert cfg.model.recompute_modules == ["moe", "core_attn", "moe_act"]
     stages = cfg.model.hybrid_layer_pattern.split("|")
     assert [len(stage) for stage in stages] == [9, 9, 9, 9, 9, 9, 9, 9, 10, 10, 11, 5]
-    assert "".join(stages) == ULTRA_HYBRID_LAYER_PATTERN
+    assert "".join(stages) == NEMOTRON_3_ULTRA_HYBRID_LAYER_PATTERN
     # A recomputed TE norm as a stage's first backward op fails on TE >= 2.18 (invalid device context).
     assert "layernorm" not in cfg.model.recompute_modules
     assert cfg.dist.distributed_timeout_minutes == 90

tests/unit_tests/recipes/recipe_test_utils.py

@@ -117,6 +117,12 @@ def exported_recipe_factory_keys(module: ModuleType) -> set[tuple[str, str]]:
     }
 
 
+# Nemotron 3 Ultra's 108-layer decoder pattern from the Hugging Face config.
+NEMOTRON_3_ULTRA_HYBRID_LAYER_PATTERN = (
+    "MEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEME"
+)
+
+
 class _OfflineModelProvider:
     """Mutable provider with the fields HF-backed recipes read before writing.
 
@@ -195,6 +201,9 @@ def to_megatron_provider(self, *args: object, **kwargs: object) -> _OfflineModel
             if model_id in {"zai-org/glm-5.2", "zai-org/glm-5.3"}:
                 provider.dsa_indexer_topk_freq = 4
                 provider.dsa_indexer_skip_topk_offset = 3
+        if model_id == "nvidia/nvidia-nemotron-3-ultra-550b-a55b-bf16":
+            provider.hybrid_layer_pattern = NEMOTRON_3_ULTRA_HYBRID_LAYER_PATTERN
+            provider.num_layers = len(provider.hybrid_layer_pattern)
         if "deepseek-v4-" in model_id:
             logical_layers = 61 if "deepseek-v4-pro" in model_id else 43
             provider.hybrid_layer_pattern = "WEWE" + "".join(

tests/unit_tests/recipes/test_nemotronh_recipes.py

@@ -27,7 +27,10 @@
 import torch
 
 from megatron.bridge.training.utils.omegaconf_utils import OverridesError, process_config_with_overrides
-from tests.unit_tests.recipes.recipe_test_utils import patch_recipe_module_global
+from tests.unit_tests.recipes.recipe_test_utils import (
+    NEMOTRON_3_ULTRA_HYBRID_LAYER_PATTERN,
+    patch_recipe_module_global,
+)
 
 
 _nemotronh_module = importlib.import_module("megatron.bridge.recipes.nemotronh")
@@ -77,10 +80,7 @@ class _FakeUltraModelProvider(_FakeModelProvider):
 
     def __init__(self) -> None:
         super().__init__()
-        self.hybrid_layer_pattern = (
-            "MEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEME"
-            "MEM*EMEMEM*EMEMEMEM*EMEMEMEM*EMEMEM*EMEMEMEM*EMEMEMEME"
-        )
+        self.hybrid_layer_pattern = NEMOTRON_3_ULTRA_HYBRID_LAYER_PATTERN
 
 
 class _FakeUltraAutoBridge(_FakeAutoBridge):