[None][fix] update skip config (NVIDIA#6891)

Signed-off-by: Ivy Zhang <[email protected]>
Signed-off-by: Wangshanshan <[email protected]>

Author: crazydemo

tests/integration/defs/accuracy/test_disaggregated_serving.py

@@ -345,6 +345,7 @@ class TestLlama3_1_8BInstruct(LlmapiAccuracyTestHarness):
     MODEL_PATH = f"{llm_models_root()}/llama-3.1-model/Llama-3.1-8B-Instruct"
 
     @pytest.mark.skip_less_device_memory(32000)
+    @pytest.mark.skip_less_device(2)
     @pytest.mark.parametrize("disable_overlap_scheduler", [False, True])
     def test_auto_dtype(self, disable_overlap_scheduler):
         ctx_server_config = {"disable_overlap_scheduler": True}
@@ -374,6 +375,8 @@ def test_auto_dtype(self, disable_overlap_scheduler):
             task = GSM8K(self.MODEL_NAME)
             task.evaluate(llm)
 
+    @pytest.mark.skip_less_device(2)
+    @skip_pre_hopper
     def test_ngram(self):
         speculative_decoding_config = {
             "decoding_type": "NGram",
@@ -424,6 +427,7 @@ def test_ngram(self):
     @skip_pre_hopper
     @parametrize_with_ids("overlap_scheduler", [True, False])
     @parametrize_with_ids("eagle3_one_model", [True, False])
+    @pytest.mark.skip_less_device(2)
     def test_eagle3(self, overlap_scheduler, eagle3_one_model):
         speculative_decoding_config = {
             "decoding_type": "Eagle",
@@ -578,7 +582,6 @@ def test_tp_pp_symmetric(self, tp, pp, testset):
         return run_parallel_test(self.MODEL_NAME, self.MODEL_PATH, pp, tp, pp,
                                  tp, 1, 1, [get_accuracy_task(testset)])
 
-    @pytest.mark.skip_less_device(4)
     @parametrize_with_ids("ctx_pp", [2, 4])
     @parametrize_with_ids("gen_tp", [1, 2])
     @pytest.mark.parametrize("testset", ["GSM8K", "MMLU"])
@@ -589,20 +592,18 @@ def test_ctx_pp_gen_tp_asymmetric(self, ctx_pp, gen_tp, testset):
         return run_parallel_test(self.MODEL_NAME, self.MODEL_PATH, ctx_pp, 1, 1,
                                  gen_tp, 1, 1, [get_accuracy_task(testset)])
 
-    @pytest.mark.skip_less_device(4)
     @pytest.mark.parametrize("testset", ["GSM8K", "MMLU"])
     def test_multi_instance(self, testset):
         return run_parallel_test(self.MODEL_NAME, self.MODEL_PATH, 1, 1, 1, 1,
                                  2, 2, [get_accuracy_task(testset)])
 
 
-@pytest.mark.skip_less_device_memory(140000)
-@pytest.mark.timeout(3600)
-@pytest.mark.skip_less_device(4)
 class TestLlama4ScoutInstruct(LlmapiAccuracyTestHarness):
     MODEL_NAME = "meta-llama/Llama-4-Scout-17B-16E-Instruct"
     MODEL_PATH = f"{llm_models_root()}/llama4-models/Llama-4-Scout-17B-16E-Instruct"
 
+    @pytest.mark.skip_less_device_memory(140000)
+    @pytest.mark.timeout(3600)
     @pytest.mark.skip_less_device(8)
     @pytest.mark.parametrize("overlap_scheduler", [False, True])
     def test_auto_dtype(self, overlap_scheduler):
@@ -683,7 +684,7 @@ def test_nixl_backend(self):
     @parametrize_with_ids("overlap_scheduler", [True, False])
     @parametrize_with_ids("mtp_nextn",
                           [0, pytest.param(2, marks=skip_pre_hopper)])
-    @pytest.mark.skip_less_device(4)
+    @pytest.mark.skip_less_device(8)
     def test_auto_dtype(self, overlap_scheduler, mtp_nextn):
         ctx_server_config = {"disable_overlap_scheduler": True}
         gen_server_config = {"disable_overlap_scheduler": not overlap_scheduler}
@@ -727,6 +728,7 @@ class TestGemma3_1BInstruct(LlmapiAccuracyTestHarness):
     MODEL_NAME = "google/gemma-3-1b-it"
     MODEL_PATH = f"{llm_models_root()}/gemma/gemma-3-1b-it/"
 
+    @pytest.mark.skip_less_device(2)
     @pytest.mark.parametrize("overlap_scheduler", [False, True])
     def test_auto_dtype(self, overlap_scheduler):
         pytest.skip(
@@ -816,8 +818,9 @@ def test_nixl_backend(self):
             task = GSM8K(self.MODEL_NAME)
             task.evaluate(llm)
 
-    @pytest.mark.parametrize("overlap_scheduler", [False, True])
     @skip_pre_hopper
+    @pytest.mark.skip_less_device(2)
+    @pytest.mark.parametrize("overlap_scheduler", [False, True])
     def test_auto_dtype(self, overlap_scheduler):
         ctx_server_config = {
             "disable_overlap_scheduler": True,

tests/integration/defs/accuracy/test_llm_api_pytorch.py

@@ -823,25 +823,26 @@ def test_auto_dtype(self):
 
 class TestMistralSmall24B(LlmapiAccuracyTestHarness):
     MODEL_NAME = "mistralai/Mistral-Small-3.1-24B-Instruct-2503"
+    MODEL_PATH = f"{llm_models_root()}/Mistral-Small-3.1-24B-Instruct-2503"
 
     @pytest.mark.skip_less_device_memory(80000)
-    @pytest.mark.parametrize(
-        "model_path, expected_quant_algo",
-        [
-            # Original bfloat16 model.
-            (f"{llm_models_root()}/Mistral-Small-3.1-24B-Instruct-2503", None),
-            # FP8 model.
-            pytest.param(
-                f"{llm_models_root()}/Mistral-Small-3.1-24B-Instruct-2503-fp8",
-                QuantAlgo.FP8,
-                marks=skip_pre_ada,
-            ),
-        ],
-    )
-    def test_auto_dtype(self, model_path, expected_quant_algo):
+    def test_auto_dtype(self):
         kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.75)
+        with LLM(self.MODEL_PATH, kv_cache_config=kv_cache_config) as llm:
+            task = CnnDailymail(self.MODEL_NAME)
+            task.evaluate(llm)
+            task = MMLU(self.MODEL_NAME)
+            task.evaluate(llm)
+            task = GSM8K(self.MODEL_NAME)
+            task.evaluate(llm)
+
+    @skip_pre_ada
+    @pytest.mark.skip_less_device_memory(80000)
+    def test_fp8(self):
+        kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.75)
+        model_path = f"{llm_models_root()}/Mistral-Small-3.1-24B-Instruct-2503-fp8"
         with LLM(model_path, kv_cache_config=kv_cache_config) as llm:
-            assert llm.args.quant_config.quant_algo == expected_quant_algo
+            assert llm.args.quant_config.quant_algo == QuantAlgo.FP8
             task = CnnDailymail(self.MODEL_NAME)
             task.evaluate(llm)
             task = MMLU(self.MODEL_NAME)

tests/integration/test_lists/qa/llm_function_full.txt

@@ -464,8 +464,8 @@ accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=False]
 accuracy/test_llm_api_pytorch.py::TestMistral7B::test_auto_dtype
 accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype
-accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_auto_dtype[/scratch.trt_llm_data/llm-models/Mistral-Small-3.1-24B-Instruct-2503-None]
-accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_auto_dtype[/scratch.trt_llm_data/llm-models/Mistral-Small-3.1-24B-Instruct-2503-fp8-FP8]
+accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_auto_dtype
+accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_fp8
 accuracy/test_llm_api_pytorch.py::TestLlama4MaverickInstruct::test_auto_dtype[tp8-cuda_graph=False]
 accuracy/test_llm_api_pytorch.py::TestLlama4MaverickInstruct::test_auto_dtype[tp8ep4-cuda_graph=True]
 accuracy/test_llm_api_pytorch.py::TestLlama4MaverickInstruct::test_auto_dtype[tp8ep8-cuda_graph=True]

tests/integration/test_lists/qa/llm_function_sanity.txt

@@ -58,6 +58,7 @@ accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-cutlass]
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-triton]
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-trtllm]
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4a16[dp4]
+accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_auto_dtype
 accuracy/test_llm_api_pytorch.py::TestKanana_Instruct::test_auto_dtype
 accuracy/test_llm_api_pytorch.py::TestKimiK2::test_fp8_blockscale[latency]
 accuracy/test_llm_api_pytorch.py::TestLlama3_1_8B::test_nvfp4

tests/integration/test_lists/test-db/l0_dgx_b200.yml

@@ -41,10 +41,10 @@ l0_dgx_b200:
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[dep4_latency_moe_cutlass-torch_compile=False]
   - accuracy/test_llm_api_pytorch.py::TestLlama4ScoutInstruct::test_fp4[tp4-cuda_graph=True]
   - disaggregated/test_disaggregated.py::test_disaggregated_deepseek_v3_lite_fp8_ucx[DeepSeek-V3-Lite-fp8]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-TRTLLM]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-CUTLASS]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-TRITON]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-TRTLLM]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-triton]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-triton]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-triton]
   - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4a16[dp4]
   - disaggregated/test_disaggregated.py::test_disaggregated_benchmark_on_diff_backends[DeepSeek-V3-Lite-bf16]
   - disaggregated/test_disaggregated.py::test_disaggregated_benchmark_on_diff_backends[llama-3.1-8b-instruct-hf-fp8]
@@ -94,11 +94,11 @@ l0_dgx_b200:
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[dep4_latency_moe_cutlass-torch_compile=True]
   - accuracy/test_llm_api_pytorch.py::TestLlama4ScoutInstruct::test_fp8[tp4-cuda_graph=True]
   - accuracy/test_disaggregated_serving.py::TestQwen3_30B_A3B::test_mixed_ctx_gen_model[ctxpp2gentp2]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-CUTLASS]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-TRITON]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-TRTLLM]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-CUTLASS]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-TRITON]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-triton]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-trtllm]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-triton]
   - disaggregated/test_disaggregated.py::test_disaggregated_benchmark_on_diff_backends[llama-v3-8b-hf]
   # ------------- AutoDeploy tests ---------------
   - accuracy/test_llm_api_autodeploy.py::TestLlama3_1_8B::test_auto_dtype

tests/integration/test_lists/test-db/l0_dgx_h100.yml

@@ -151,12 +151,12 @@ l0_dgx_h100:
       backend: pytorch
       auto_trigger: gpt_oss
   tests:
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-CUTLASS]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-TRITON]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-CUTLASS]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-TRITON]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-CUTLASS]
-  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-TRITON]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-triton]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-triton]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-triton]
   - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4a16[dp4]
 - condition:
     ranges:

tests/integration/test_lists/test-db/l0_h100.yml

@@ -223,8 +223,8 @@ l0_h100:
   - accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_fp8_prequantized
   - accuracy/test_llm_api_pytorch.py::TestGemma3_27BInstruct::test_fp8_prequantized
   - accuracy/test_llm_api_pytorch.py::TestGemma3_27BInstruct::test_auto_dtype
-  - accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_auto_dtype[/scratch.trt_llm_data/llm-models/Mistral-Small-3.1-24B-Instruct-2503-None]
-  - accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_auto_dtype[/scratch.trt_llm_data/llm-models/Mistral-Small-3.1-24B-Instruct-2503-fp8-FP8]
+  - accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_auto_dtype
+  - accuracy/test_llm_api_pytorch.py::TestMistralSmall24B::test_fp8
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_fp8_block_scales[latency-torch_compile=False]
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_fp8_block_scales[latency-torch_compile=True]
   - accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_guided_decoding[llguidance]