[Bug]: DGX Spark playbook supported models fail in trtllm-serve: Nemotron-3 Nano Omni and GPT-OSS rejected as unsupported due to MODEL_MAP registry mismatch

### System Info Environment ----------- DGX Spark (128 GB) Container: nvcr.io/nvidia/tensorrt-llm/release:1.3.0rc13 Also reproduced on: nvcr.io/nvidia/tensorrt-llm/release:1.3.0rc20 Models ------ nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4 openai/gpt-oss-20b Command ------- trtllm-serve serve \ --backend tensorrt \ --trust_remote_code \ nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4 Expected -------- Model is supported according to the DGX Spark playbook. Actual ------ NotImplementedError: The given huggingface model architecture NemotronH_Nano_Omni_Reasoning_V3 is not supported in TRT-LLM yet Investigation ------------- - AutoConfig.from_pretrained(..., trust_remote_code=True) succeeds. - Architecture returned: NemotronH_Nano_Omni_Reasoning_V3 - The architecture is registered in: tensorrt_llm._torch.models.modeling_utils.MODEL_CLASS_MAPPING - trtllm-serve / llm_utils imports: tensorrt_llm.models.automodel.AutoModelForCausalLM - AutoModelForCausalLM.get_trtllm_model_class() looks up MODEL_MAP. - MODEL_MAP does not contain NemotronH_Nano_Omni_Reasoning_V3. - This appears to be a registry mismatch between the legacy automodel path and the newer _torch model registry. Question -------- Is the DGX Spark playbook intended to use a different code path than trtllm-serve, or is this a regression in the released containers? ### Who can help? trtllm-serve serve \ --backend tensorrt \ --trust_remote_code \ --host 0.0.0.0 \ --port 8000 \ nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4 ---> Identical command for gpt-oss <--- ### Information - [ ] The official example scripts - [ ] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...) - [ ] My own task or dataset (give details below) ### Reproduction Launch the official TensorRT-LLM container from the DGX Spark playbook (reproduced on both 1.3.0rc13 and 1.3.0rc20). Set a valid Hugging Face token: export HF_TOKEN=<your_token> Start the container with GPU access: docker run --rm -it \ --gpus all \ --ipc=host \ --network=host \ --ulimit memlock=-1 \ --ulimit stack=67108864 \ -e HF_TOKEN=$HF_TOKEN \ -v $HOME/.cache/huggingface:/root/.cache/huggingface \ nvcr.io/nvidia/tensorrt-llm/release:1.3.0rc20 \ bash Verify the TensorRT-LLM version: python3 -c "import tensorrt_llm; print(tensorrt_llm.__version__)" Launch a model listed as supported in the DGX Spark playbook: trtllm-serve serve \ --backend tensorrt \ --trust_remote_code \ --host 0.0.0.0 \ --port 8000 \ nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4 Observe the failure: NotImplementedError: The given huggingface model architecture NemotronH_Nano_Omni_Reasoning_V3 is not supported in TRT-LLM yet Repeat with another model listed as supported (e.g. openai/gpt-oss-20b) and observe a similar architecture-not-supported failure. ### Expected behavior The models should serve properly. ### actual behavior NotImplementedError: The given huggingface model architecture NemotronH_Nano_Omni_Reasoning_V3 is not supported in TRT-LLM yet ### additional notes I get this: Skipping import of cpp extensions due to incompatible torch version 2.11.0a0+eb65b36914.nv26.02 for torchao version 0.15.0 Please see https://github.com/pytorch/ao/issues/2919 for more info /usr/local/lib/python3.12/dist-packages/modelopt/torch/__init__.py:36: UserWarning: transformers version 5.5.4 is incompatible with nvidia-modelopt and may cause issues. Please install recommended version with `pip install nvidia-modelopt[hf]` if working with HF models. _warnings.warn( [TensorRT-LLM] TensorRT LLM version: 1.3.0rc20 /usr/local/lib/python3.12/dist-packages/tensorrt_llm/serve/openai_protocol.py:145: UserWarning: Field name "schema" in "ResponseFormat" shadows an attribute in parent "OpenAIBaseModel" class ResponseFormat(OpenAIBaseModel): [07/05/2026-01:22:02] [TRT-LLM] [I] [llmapi] Using LLM with TensorRT backend [07/05/2026-01:22:02] [TRT-LLM] [I] [plugin] Set PluginConfig.nccl_plugin to None. Loading Model: [1/3] Downloading HF model [07/05/2026-01:22:02] [TRT-LLM] [I] [llmapi] Downloading model nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4 from HuggingFace [07/05/2026-01:22:02] [TRT-LLM] [I] [llmapi] Finished downloading model nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4 from HuggingFace Downloaded model to /root/.cache/huggingface/hub/models--nvidia--Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4/snapshots/dc5f0b0bfddf8b6e0f5891475be9af05b80126fe Time: 0.310s Loading Model: [2/3] Loading HF model to memory Traceback (most recent call last): File "/usr/local/bin/trtllm-serve", line 6, in <module> sys.exit(main()) ^^^^^^ File "/usr/local/lib/python3.12/dist-packages/click/core.py", line 1485, in __call__ return self.main(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/click/core.py", line 1406, in main rv = self.invoke(ctx) ^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/click/core.py", line 1873, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/click/core.py", line 1269, in invoke return ctx.invoke(self.callback, **ctx.params) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/click/core.py", line 824, in invoke return callback(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/commands/serve.py", line 1154, in serve _serve_llm() File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/commands/serve.py", line 1126, in _serve_llm launch_server( File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/commands/serve.py", line 385, in launch_server llm = LLM(**llm_args) ^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm.py", line 1519, in __init__ super().__init__(model, tokenizer, tokenizer_mode, skip_tokenizer_init, File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm.py", line 349, in __init__ self._build_model() File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm.py", line 1553, in _build_model super()._build_model() File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm.py", line 1381, in _build_model self._engine_dir, self._hf_model_dir = model_loader() ^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm_utils.py", line 818, in __call__ return self._build_model(), self._hf_model_dir ^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm_utils.py", line 959, in _build_model build_task(self.get_engine_dir()) File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm_utils.py", line 913, in build_task model_loader(engine_dir=engine_dir) File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm_utils.py", line 265, in __call__ pipeline() File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm_utils.py", line 206, in __call__ self.step_forward() File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm_utils.py", line 235, in step_forward self.step_handlers[self.counter]() File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/llmapi/llm_utils.py", line 529, in _load_model_from_hf model_cls = AutoModelForCausalLM.get_trtllm_model_class( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/tensorrt_llm/models/automodel.py", line 85, in get_trtllm_model_class raise NotImplementedError( NotImplementedError: The given huggingface model architecture NemotronH_Nano_Omni_Reasoning_V3 is not supported in TRT-LLM yet ### Before submitting a new issue... - [x] Make sure you already searched for relevant issues, and checked the [documentation](https://nvidia.github.io/TensorRT-LLM/) and [examples](https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples) for answers to frequently asked questions.