Support for NemotronH Nano VLM (#23644)

Signed-off-by: Daniel Afrimi <danielafrimi8@gmail.com>

tests/models/registry.py

@@ -515,6 +515,9 @@ _MULTIMODAL_EXAMPLE_MODELS = {
                               trust_remote_code=True),
     "Llama_Nemotron_Nano_VL" : _HfExamplesInfo("nvidia/Llama-3.1-Nemotron-Nano-VL-8B-V1", # noqa: E501
                                                      trust_remote_code=True),
+    "NemotronH_Nano_VL": _HfExamplesInfo("nano_vl_dummy",
+                                          is_available_online=False,
+                                          trust_remote_code=True),
     "Ovis": _HfExamplesInfo("AIDC-AI/Ovis2-1B", trust_remote_code=True,
                             max_transformers_version="4.53",
                             transformers_version_reason="HF model is not compatible",  # noqa: E501

vllm/config/__init__.py

@@ -1552,7 +1552,7 @@ class ModelConfig:
                        for bc in block_configs[start:end])
         else:
             # Hybrid model Jamba
-            layers_block_type_value = getattr(self.hf_config,
+            layers_block_type_value = getattr(self.hf_text_config,
                                               "layers_block_type", None)
             if layers_block_type_value is not None:
                 if hasattr(self.hf_text_config,

vllm/model_executor/models/registry.py

@@ -223,6 +223,7 @@ _MULTIMODAL_MODELS = {
     "GraniteSpeechForConditionalGeneration": ("granite_speech", "GraniteSpeechForConditionalGeneration"),  # noqa: E501
     "H2OVLChatModel": ("h2ovl", "H2OVLChatModel"),
     "InternVLChatModel": ("internvl", "InternVLChatModel"),
+    "NemotronH_Nano_VL": ("nano_nemotron_vl", "NemotronH_Nano_VL"),
     "InternS1ForConditionalGeneration": ("interns1", "InternS1ForConditionalGeneration"),  # noqa: E501
     "InternVLForConditionalGeneration": ("interns1", "InternS1ForConditionalGeneration"),  # noqa: E501
     "Idefics3ForConditionalGeneration":("idefics3","Idefics3ForConditionalGeneration"),