add gptoss20b and fix nemo nano

Signed-off-by: list <58580514+farazkh80@users.noreply.github.com>

Author: farazkh80

tests/integration/defs/accuracy/test_llm_api_pytorch.py

@@ -3958,6 +3958,34 @@ def test_w4_1gpu(self, kv_cache_dtype, moe_backend, cuda_graph,
             task.evaluate(llm,
                           extra_evaluator_kwargs=self.extra_evaluator_kwargs)
 
+    # on spark 120b accuracy takes 2.2 hours, so we do 20b for now
+    def test_w4_1gpu_20b_spark(self, mocker):
+        mocker.patch.object(GSM8K, "MAX_OUTPUT_LEN", 8192)
+        mocker.patch.dict(GSM8K.EVALUATE_KWARGS,
+                          {"scores_filter": "exact_match,flexible-extract"})
+
+        pytorch_config = dict(
+            disable_overlap_scheduler=False,
+            cuda_graph_config=CudaGraphConfig())
+
+        kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.5,
+                                        dtype="auto")
+
+        model_path = f"{llm_models_root()}/gpt_oss/gpt-oss-20b"
+        llm = LLM(model_path,
+                  tensor_parallel_size=1,
+                  pipeline_parallel_size=1,
+                  moe_expert_parallel_size=1,
+                  kv_cache_config=kv_cache_config,
+                  **pytorch_config,
+                  moe_config=MoeConfig(backend="CUTLASS"))
+
+        with llm:
+            model_name = "GPT-OSS/20B-MXFP4"
+            task = GSM8K(model_name)
+            task.evaluate(llm,
+                          extra_evaluator_kwargs=self.extra_evaluator_kwargs)
+
     def test_dummy_load_format(self):
         llm = LLM(
             self.MODEL_PATH,

tests/integration/defs/test_e2e.py

@@ -1923,7 +1923,7 @@ def test_ptp_quickstart(llm_root, llm_venv):
 def test_ptp_quickstart_advanced(llm_root, llm_venv, model_name, model_path):
     print(f"Testing {model_name}.")
     example_root = Path(os.path.join(llm_root, "examples", "llm-api"))
-    if model_name == "Nemotron-H-8B":
+    if model_name in ("Nemotron-H-8B", "Nemotron-Nano-v2-nvfp4"):
         llm_venv.run_cmd([
             str(example_root / "quickstart_advanced.py"),
             "--disable_kv_cache_reuse",

tests/integration/test_lists/test-db/l0_gb10.yml

@@ -15,9 +15,9 @@ l0_gb10:
       backend: pytorch
   tests:
   # ------------- PyTorch tests ---------------
-  - unittest/_torch/modeling -k "modeling_mllama"
-  - unittest/_torch/modeling -k "modeling_out_of_tree"
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu[True-True-cutlass-auto]
+  - accuracy/test_llm_api_pytorch.py::TestLlama3_1_8B::test_auto_dtype
+  - accuracy/test_llm_api_pytorch.py::TestLlama3_1_8B::test_nvfp4
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu_20b_spark
 - condition:
     ranges:
       system_gpu_count:
@@ -39,6 +39,7 @@ l0_gb10:
   - test_e2e.py::test_ptp_quickstart_advanced[Llama3.1-8B-FP8-llama-3.1-model/Llama-3.1-8B-Instruct-FP8]
   - test_e2e.py::test_ptp_quickstart_advanced[Llama3.1-8B-NVFP4-nvfp4-quantized/Meta-Llama-3.1-8B]
   - test_e2e.py::test_ptp_quickstart_advanced[GPT-OSS-20B-gpt_oss/gpt-oss-20b]
+  - test_e2e.py::test_ptp_quickstart_advanced[GPT-OSS-120B-gpt_oss/gpt-oss-120b]
   - test_e2e.py::test_ptp_quickstart_advanced[Qwen3-8b-fp8-Qwen3/nvidia-Qwen3-8B-FP8]
   - test_e2e.py::test_ptp_quickstart_advanced[Qwen3-8b-nvfp4-Qwen3/nvidia-Qwen3-8B-NVFP4]
   - test_e2e.py::test_ptp_quickstart_advanced[Qwen3-14b-fp8-Qwen3/nvidia-Qwen3-14B-FP8]