ci: frontload all CPU unit tests

.github/workflows/community-cpu.yml

@@ -110,54 +110,25 @@ jobs:
 
   unit:
     name: Community CPU / Unit / C++ and Python
-    if: ${{ !cancelled() }}
-    needs: ownership-impact
     runs-on: ubuntu-24.04
     timeout-minutes: 45
     permissions:
       contents: read
     steps:
-      - name: Require successful impact analysis
-        if: ${{ needs.ownership-impact.result != 'success' }}
-        env:
-          IMPACT_RESULT: ${{ needs.ownership-impact.result }}
-        run: |
-          echo "::error::Ownership and impact did not succeed: $IMPACT_RESULT"
-          exit 1
-
-      - name: Explain an empty unit scope
-        if: >-
-          needs.ownership-impact.result == 'success' &&
-          needs.ownership-impact.outputs.run_unit_tests != 'true'
-        run: echo "No source-only C++ or Python unit tests are required for this change."
-
       - name: Check out the exact PR merge
-        if: >-
-          needs.ownership-impact.result == 'success' &&
-          needs.ownership-impact.outputs.run_unit_tests == 'true'
         uses: actions/checkout@d23441a48e516b6c34aea4fa41551a30e30af803
         with:
           ref: ${{ github.sha }}
           fetch-depth: 0
           persist-credentials: false
 
       - name: Set up Python
-        if: >-
-          needs.ownership-impact.result == 'success' &&
-          needs.ownership-impact.outputs.run_unit_tests == 'true'
         uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405
         with:
           python-version: "3.12"
 
       - name: Run hardened source-only units
-        if: >-
-          needs.ownership-impact.result == 'success' &&
-          needs.ownership-impact.outputs.run_unit_tests == 'true'
-        env:
-          TRTMC_UNIT_SCOPE: ${{ needs.ownership-impact.outputs.unit_scope }}
-          TRTMC_PREMERGE_PYTHON_TEST_TARGETS: >-
-            ${{ needs.ownership-impact.outputs.python_test_targets }}
-        run: python3 -m tools.community_ci unit --scope "$TRTMC_UNIT_SCOPE"
+        run: python3 -m tools.community_ci unit --scope all
 
   required:
     name: Community CPU / Required

CMakeLists.txt

@@ -722,6 +722,7 @@ if(TRTMC_BUILD_TESTS)
 enable_testing()
 add_custom_target(trtmc_cpp_tests)
 add_custom_target(trtmc_platform_cpp_tests)
+add_custom_target(trtmc_cpu_cpp_tests)
 
 # Generic capsule test doubles. The good and wrong-identity DSOs intentionally
 # have the same filename in different directories so tests can prove that the
@@ -784,8 +785,8 @@ set_target_properties(trtmc_test_optimized_embedding PROPERTIES
 # Usage:
 #   trtmc_add_test(test_name)               -- links trtmc_core, adds src/ include
 #   trtmc_add_test(test_name REQUIRES_TRT)  -- also links TensorRT test deps
-#   trtmc_add_test(test_name MODEL_OWNED)    -- excludes it from platform units
-#   trtmc_add_test(test_name REQUIRES_GPU)   -- excludes it from CPU platform units
+#   trtmc_add_test(test_name MODEL_OWNED)    -- labels model ownership independently
+#   trtmc_add_test(test_name REQUIRES_GPU)   -- excludes it from all CPU unit targets
 function(trtmc_add_test TEST_NAME)
   cmake_parse_arguments(
     ARG
@@ -794,6 +795,10 @@ function(trtmc_add_test TEST_NAME)
     "EXTRA_INCLUDES"
     ${ARGN}
   )
+  if(ARG_UNPARSED_ARGUMENTS)
+    message(FATAL_ERROR
+      "Unknown trtmc_add_test option(s) for '${TEST_NAME}': ${ARG_UNPARSED_ARGUMENTS}")
+  endif()
   if(ARG_REQUIRES_TRT AND NOT _trtmc_can_build_trt_backend)
     message(STATUS "Skipping ${TEST_NAME}: TensorRT SDK not available")
     return()
@@ -830,13 +835,23 @@ function(trtmc_add_test TEST_NAME)
   endif()
   add_test(NAME ${TEST_NAME} COMMAND ${TEST_NAME})
   if(ARG_MODEL_OWNED)
-    set_tests_properties(${TEST_NAME} PROPERTIES LABELS "model")
-  elseif(ARG_REQUIRES_GPU)
-    set_tests_properties(${TEST_NAME} PROPERTIES LABELS "gpu")
+    set(_trtmc_test_owner_label "model")
+  else()
+    set(_trtmc_test_owner_label "platform")
+  endif()
+  if(ARG_REQUIRES_GPU)
+    set(_trtmc_test_resource_label "gpu")
   else()
+    set(_trtmc_test_resource_label "cpu")
+    add_dependencies(trtmc_cpu_cpp_tests ${TEST_NAME})
+  endif()
+  if(NOT ARG_MODEL_OWNED AND NOT ARG_REQUIRES_GPU)
     add_dependencies(trtmc_platform_cpp_tests ${TEST_NAME})
-    set_tests_properties(${TEST_NAME} PROPERTIES LABELS "platform")
   endif()
+  set_tests_properties(
+    ${TEST_NAME}
+    PROPERTIES LABELS "${_trtmc_test_owner_label};${_trtmc_test_resource_label}"
+  )
 endfunction()
 
 function(_trtmc_manifest_csv_to_list VALUE OUTPUT_VAR)
@@ -931,7 +946,11 @@ if(_trtmc_python)
   )
   add_custom_target(test_optimized_runtime_bundle_contract
     DEPENDS test_optimized_runtime_host trtmc_test_optimized_runtime)
-  set_tests_properties(test_optimized_runtime_bundle_contract PROPERTIES LABELS "platform")
+  add_dependencies(trtmc_cpu_cpp_tests test_optimized_runtime_bundle_contract)
+  set_tests_properties(
+    test_optimized_runtime_bundle_contract
+    PROPERTIES LABELS "platform;cpu"
+  )
 endif()
 # C ABI
 trtmc_add_test(test_c_abi_entry                NO_SRC_INCLUDE)

Dockerfile.community-cpu

@@ -30,6 +30,7 @@ RUN apt-get update && \
       python3.12-dev \
       python3.12-venv \
       libcublas-dev-13-3 \
+      libcurand-dev-13-3 \
       "libnvinfer-dev=${TENSORRT_APT_VERSION}" \
       "libnvinfer-headers-dev=${TENSORRT_APT_VERSION}" \
       "libnvinfer-headers-plugin-dev=${TENSORRT_APT_VERSION}" \

python/tensorrt_model_connect/families/minimax_h3/tests/test_trt_builders.py

@@ -130,6 +130,17 @@ def capture(_config, supplied, *, default_bytes):
         observed.update(supplied=supplied, default_bytes=default_bytes)
         raise WorkspaceConfigured
 
+    class FakeBuilder:
+        @staticmethod
+        def create_network(_flags):
+            return object()
+
+        @staticmethod
+        def create_builder_config():
+            return object()
+
+    monkeypatch.setattr(trt, "Builder", lambda _logger: FakeBuilder())
+    monkeypatch.setattr(op, "configure_builder", lambda _config: None)
     monkeypatch.setattr(op, "configure_workspace", capture)
     kwargs = {"workspace_bytes": workspace_bytes}
     if builder is build_text_encoder_engine:
@@ -303,6 +314,8 @@ def test_native_linear_serializes_checkpoint_bf16_without_fp32_constant() -> Non
     assert plan
 
 
+@pytest.mark.gpu
+@pytest.mark.trt
 def test_native_network_contract_counts_iattention_and_fails_closed() -> None:
     logger = trt.Logger(trt.Logger.WARNING)
     builder = trt.Builder(logger)

python/tensorrt_model_connect/families/sana_wm/tests/test_family.py

@@ -42,14 +42,25 @@
 
 
 @pytest.fixture(autouse=True)
-def _clear_sana_wm_env(monkeypatch) -> None:
+def _clear_sana_wm_env(monkeypatch, tmp_path) -> None:
     for name in _SANA_WM_ENV_VARS:
         monkeypatch.delenv(name, raising=False)
     monkeypatch.setattr(
         sana_wm_plugin_mod,
         "_cached_stage1_text_encoder_dir",
         lambda _raw_config: None,
     )
+    native_plugin = tmp_path / "libtrtmc_sana_wm_native_plugin.so"
+    native_plugin.write_bytes(b"CPU unit native plugin fixture")
+    native_plugin_builder = importlib.import_module(
+        "tensorrt_model_connect.families.sana_wm.native_plugin_builder"
+    )
+    monkeypatch.setattr(
+        native_plugin_builder,
+        "ensure_native_plugin",
+        lambda **_kwargs: native_plugin,
+    )
+    monkeypatch.setattr(sana_wm_plugin_mod.ctypes, "CDLL", lambda *_args, **_kwargs: object())
 
 
 def _sana_yaml() -> str:

requirements/community-ci.txt

@@ -11,3 +11,4 @@ pytest==8.4.2
 pytest-xdist==3.8.0
 jsonschema==4.26.0
 Pillow==12.2.0
+pyarrow==25.0.1

src/runtime/models/albert/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_albert.so"
 runtime_plugins = ["plugin.cpp|register_albert_plugin"]
 runtime_strategies = ["albert_encoder_only"]
 runtime_tests = [
-  "test_albert_encoder_pipeline|test_albert_encoder_pipeline.cpp|trtmc_model_albert,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_albert_encoder_pipeline|test_albert_encoder_pipeline.cpp|trtmc_model_albert,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/bark/MODEL.toml

@@ -10,5 +10,5 @@ legacy_runtime_strategy_aliases = ["text_to_audio|default|_|_|text_to_audio_bark
 runtime_tests = [
   "test_bark_config_schema|test_bark_config_schema.cpp|trtmc_model_bark|_|_",
   "test_bark_generation_plan|test_bark_generation_plan.cpp|_|_|_",
-  "test_bark_pipeline|test_bark_pipeline.cpp|trtmc_model_bark|_|REQUIRES_TRT",
+  "test_bark_pipeline|test_bark_pipeline.cpp|trtmc_model_bark|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/bert/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_bert.so"
 runtime_plugins = ["plugin.cpp|register_bert_plugin"]
 runtime_strategies = ["bert_encoder_only", "bert_embedding"]
 runtime_tests = [
-  "test_bert_encoder_pipeline|test_bert_encoder_pipeline.cpp|trtmc_model_bert,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_bert_encoder_pipeline|test_bert_encoder_pipeline.cpp|trtmc_model_bert,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/canary/MODEL.toml

@@ -9,5 +9,5 @@ runtime_tests = [
   "test_canary_mel_spectrogram|test_canary_mel_spectrogram.cpp|_|canary_mel_spectrogram.cpp|_",
   "test_canary_host_plan|test_canary_host_plan.cpp|_|_|_",
   "test_canary_decode_policy|test_canary_decode_policy.cpp|_|_|_",
-  "test_canary_pipeline|test_canary_pipeline.cpp|trtmc_model_canary|_|REQUIRES_TRT",
+  "test_canary_pipeline|test_canary_pipeline.cpp|trtmc_model_canary|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/convbert/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_convbert.so"
 runtime_plugins = ["plugin.cpp|register_convbert_plugin"]
 runtime_strategies = ["convbert_encoder_only"]
 runtime_tests = [
-  "test_convbert_encoder_pipeline|test_convbert_encoder_pipeline.cpp|trtmc_model_convbert,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_convbert_encoder_pipeline|test_convbert_encoder_pipeline.cpp|trtmc_model_convbert,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/deberta/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_deberta.so"
 runtime_plugins = ["plugin.cpp|register_deberta_plugin"]
 runtime_strategies = ["deberta_encoder_only"]
 runtime_tests = [
-  "test_deberta_encoder_pipeline|test_deberta_encoder_pipeline.cpp|trtmc_model_deberta,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_deberta_encoder_pipeline|test_deberta_encoder_pipeline.cpp|trtmc_model_deberta,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/deepseek_ocr/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_deepseek_ocr.so"
 runtime_plugins = ["plugin.cpp|register_deepseek_ocr_plugin"]
 runtime_strategies = ["deepseek_ocr_vision_language"]
 runtime_tests = [
-  "test_deepseek_ocr_vl_pipeline|test_deepseek_ocr_vl_pipeline.cpp|trtmc_model_deepseek_ocr,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_deepseek_ocr_vl_pipeline|test_deepseek_ocr_vl_pipeline.cpp|trtmc_model_deepseek_ocr,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/distilbert/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_distilbert.so"
 runtime_plugins = ["plugin.cpp|register_distilbert_plugin"]
 runtime_strategies = ["distilbert_encoder_only"]
 runtime_tests = [
-  "test_distilbert_encoder_pipeline|test_distilbert_encoder_pipeline.cpp|trtmc_model_distilbert,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_distilbert_encoder_pipeline|test_distilbert_encoder_pipeline.cpp|trtmc_model_distilbert,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/dpr/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_dpr.so"
 runtime_plugins = ["plugin.cpp|register_dpr_plugin"]
 runtime_strategies = ["dpr_encoder_only"]
 runtime_tests = [
-  "test_dpr_encoder_pipeline|test_dpr_encoder_pipeline.cpp|trtmc_model_dpr,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_dpr_encoder_pipeline|test_dpr_encoder_pipeline.cpp|trtmc_model_dpr,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/eagle_vlm/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_eagle_vlm.so"
 runtime_plugins = ["plugin.cpp|register_eagle_vlm_plugin"]
 runtime_strategies = ["eagle_vlm_embedding", "eagle_vlm_reranking"]
 runtime_tests = [
-  "test_eagle_vlm_encoder_pipeline|test_eagle_vlm_encoder_pipeline.cpp|trtmc_model_eagle_vlm,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_eagle_vlm_encoder_pipeline|test_eagle_vlm_encoder_pipeline.cpp|trtmc_model_eagle_vlm,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/electra/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_electra.so"
 runtime_plugins = ["plugin.cpp|register_electra_plugin"]
 runtime_strategies = ["electra_encoder_only"]
 runtime_tests = [
-  "test_electra_encoder_pipeline|test_electra_encoder_pipeline.cpp|trtmc_model_electra,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_electra_encoder_pipeline|test_electra_encoder_pipeline.cpp|trtmc_model_electra,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/elf_flow/MODEL.toml

@@ -7,6 +7,6 @@ runtime_plugins = ["plugin.cpp|register_elf_flow_plugin"]
 runtime_strategies = ["elf_flow"]
 runtime_tests = [
   "test_elf_flow_pipeline|test_elf_flow_pipeline.cpp|trtmc_model_elf_flow|_|_",
-  "test_elf_flow_sampling_kernels|test_elf_flow_sampling_kernels.cpp|trtmc_model_elf_flow|_|REQUIRES_TRT",
+  "test_elf_flow_sampling_kernels|test_elf_flow_sampling_kernels.cpp|trtmc_model_elf_flow|_|REQUIRES_TRT,REQUIRES_GPU",
 ]
 task_strategy = "diffusion_text_generation"

src/runtime/models/flux/MODEL.toml

@@ -8,7 +8,8 @@ runtime_strategies = ["diffusion_flux"]
 legacy_runtime_strategy_aliases = ["diffusion|contains|diffusion_backend_type|flux|diffusion_flux"]
 runtime_link_libraries = ["cublas"]
 runtime_tests = [
-  "test_flux_pipeline|test_flux_pipeline.cpp|trtmc_model_flux|_|_",
+  "test_flux_host_contracts|test_flux_host_contracts.cpp|_|_|_",
+  "test_flux_pipeline|test_flux_pipeline.cpp|trtmc_model_flux|_|REQUIRES_GPU",
   "test_flux_generation_plan|test_flux_generation_plan.cpp|_|_|_",
   "test_flux_denoising_step_seam|test_flux_denoising_step_seam.cpp|_|_|_",
   "test_flux_batch_utils|test_flux_batch_utils.cpp|_|_|_",

src/runtime/models/fnet/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_fnet.so"
 runtime_plugins = ["plugin.cpp|register_fnet_plugin"]
 runtime_strategies = ["fnet_encoder_only"]
 runtime_tests = [
-  "test_fnet_encoder_pipeline|test_fnet_encoder_pipeline.cpp|trtmc_model_fnet,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_fnet_encoder_pipeline|test_fnet_encoder_pipeline.cpp|trtmc_model_fnet,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/internlm/MODEL.toml

@@ -5,10 +5,10 @@ id = "internlm"
 runtime_library = "libtrtmc_model_internlm.so"
 runtime_plugins = ["plugin.cpp|register_internlm_plugin"]
 runtime_strategies = ["internlm_decoder_kv_cache"]
-[validation_profiles]
-decoder_debug = ["internlm_decoder_kv_cache"]
-
 runtime_tests = [
   "test_internlm_chat_template|test_internlm_chat_template.cpp|trtmc_model_internlm|_|_",
   "test_internlm_pipeline|test_internlm_pipeline.cpp|trtmc_model_internlm|_|REQUIRES_GPU",
 ]
+
+[validation_profiles]
+decoder_debug = ["internlm_decoder_kv_cache"]

src/runtime/models/lance/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_lance.so"
 runtime_plugins = ["plugin.cpp|register_lance_plugin"]
 runtime_strategies = ["lance_vision_language"]
 runtime_tests = [
-  "test_lance_vl_pipeline|test_lance_vl_pipeline.cpp|trtmc_model_lance,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_lance_vl_pipeline|test_lance_vl_pipeline.cpp|trtmc_model_lance,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/llama/MODEL.toml

@@ -5,12 +5,12 @@ id = "llama"
 runtime_library = "libtrtmc_model_llama.so"
 runtime_plugins = ["plugin.cpp|register_llama_plugin"]
 runtime_strategies = ["llama_decoder_kv_cache"]
-[validation_profiles]
-decoder_debug = ["llama_decoder_kv_cache"]
-
 runtime_tests = [
   "test_llama_plugin_helpers|test_llama_plugin_helpers.cpp|_|plugin_helpers.cpp|_",
   "test_llama_chat_template|test_llama_chat_template.cpp|trtmc_model_llama|_|_",
-  "test_llama_pipeline|test_llama_pipeline.cpp|trtmc_model_llama,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_llama_pipeline|test_llama_pipeline.cpp|trtmc_model_llama,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
   "test_llama_native_kv_cache|test_llama_native_kv_cache.cpp|trtmc_model_llama|_|REQUIRES_GPU",
 ]
+
+[validation_profiles]
+decoder_debug = ["llama_decoder_kv_cache"]

src/runtime/models/magpie/MODEL.toml

@@ -15,5 +15,5 @@ runtime_tests = [
   "test_magpie_generation_plan|test_magpie_generation_plan.cpp|_|_|_",
   "test_magpie_text_completion_policy|test_magpie_text_completion_policy.cpp|_|_|_",
   "test_magpie_codec_plan|test_magpie_codec_plan.cpp|_|_|_",
-  "test_magpie_pipeline|test_magpie_pipeline.cpp|trtmc_model_magpie|_|REQUIRES_TRT",
+  "test_magpie_pipeline|test_magpie_pipeline.cpp|trtmc_model_magpie|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/mamba/MODEL.toml

@@ -5,10 +5,10 @@ id = "mamba"
 runtime_library = "libtrtmc_model_mamba.so"
 runtime_plugins = ["plugin.cpp|register_mamba_plugin"]
 runtime_strategies = ["mamba_ssm_recurrent"]
-[validation_profiles]
-decoder_debug = ["mamba_ssm_recurrent"]
-
 runtime_tests = [
   "test_mamba_recurrent_output_initializers|test_mamba_recurrent_output_initializers.cpp|_|_|_",
-  "test_mamba_recurrent_pipeline|test_mamba_recurrent_pipeline.cpp|trtmc_model_mamba,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_mamba_recurrent_pipeline|test_mamba_recurrent_pipeline.cpp|trtmc_model_mamba,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]
+
+[validation_profiles]
+decoder_debug = ["mamba_ssm_recurrent"]

src/runtime/models/modernbert/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_modernbert.so"
 runtime_plugins = ["plugin.cpp|register_modernbert_plugin"]
 runtime_strategies = ["modernbert_encoder_only"]
 runtime_tests = [
-  "test_modernbert_encoder_pipeline|test_modernbert_encoder_pipeline.cpp|trtmc_model_modernbert,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_modernbert_encoder_pipeline|test_modernbert_encoder_pipeline.cpp|trtmc_model_modernbert,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/mpnet/MODEL.toml

@@ -6,5 +6,5 @@ runtime_library = "libtrtmc_model_mpnet.so"
 runtime_plugins = ["plugin.cpp|register_mpnet_plugin"]
 runtime_strategies = ["mpnet_encoder_only"]
 runtime_tests = [
-  "test_mpnet_encoder_pipeline|test_mpnet_encoder_pipeline.cpp|trtmc_model_mpnet,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_mpnet_encoder_pipeline|test_mpnet_encoder_pipeline.cpp|trtmc_model_mpnet,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]

src/runtime/models/nemotron_h/MODEL.toml

@@ -5,11 +5,11 @@ id = "nemotron_h"
 runtime_library = "libtrtmc_model_nemotron_h.so"
 runtime_plugins = ["plugin.cpp|register_nemotron_h_plugin"]
 runtime_strategies = ["nemotron_h_hybrid_mamba_attention"]
-[validation_profiles]
-decoder_debug = ["nemotron_h_hybrid_mamba_attention"]
-
 runtime_tests = [
   "test_nemotron_h_backend_cuda_graph|test_nemotron_h_backend_cuda_graph.cpp|trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
   "test_nemotron_h_recurrent_output_initializers|test_nemotron_h_recurrent_output_initializers.cpp|_|_|_",
-  "test_nemotron_h_recurrent_pipeline|test_nemotron_h_recurrent_pipeline.cpp|trtmc_model_nemotron_h,trtmc_backend_trt|_|REQUIRES_TRT",
+  "test_nemotron_h_recurrent_pipeline|test_nemotron_h_recurrent_pipeline.cpp|trtmc_model_nemotron_h,trtmc_backend_trt|_|REQUIRES_TRT,REQUIRES_GPU",
 ]
+
+[validation_profiles]
+decoder_debug = ["nemotron_h_hybrid_mamba_attention"]

src/runtime/models/nemotron_labs_diffusion/MODEL.toml

@@ -5,9 +5,9 @@ id = "nemotron_labs_diffusion"
 runtime_library = "libtrtmc_model_nemotron_labs_diffusion.so"
 runtime_plugins = ["plugin.cpp|register_nemotron_labs_diffusion_plugin"]
 runtime_strategies = ["nemotron_labs_diffusion"]
-[validation_profiles]
-decoder_debug = ["nemotron_labs_diffusion"]
-
 runtime_tests = [
   "test_nemotron_labs_diffusion_chat_template|test_nemotron_labs_diffusion_chat_template.cpp|trtmc_model_nemotron_labs_diffusion|_|_",
 ]
+
+[validation_profiles]
+decoder_debug = ["nemotron_labs_diffusion"]