[https://nvbugs/6222480][test] fix stress test issue on H100

tests/integration/defs/disaggregated/test_configs/disagg_config_ctxtp2_gentp2_gptoss_tllm_eagle.yaml → tests/integration/defs/disaggregated/test_configs/disagg_config_ctxtp2_gentp2_gptoss_eagle_triton.yaml

@@ -22,7 +22,7 @@ context_servers:
     enable_block_reuse: false
     free_gpu_memory_fraction: 0.8
   moe_config:
-    backend: CUTLASS
+    backend: TRITON
   cuda_graph_config: null
   print_iter_log: true
   cache_transceiver_config:
@@ -45,7 +45,7 @@ generation_servers:
     enable_block_reuse: false
     free_gpu_memory_fraction: 0.8
   moe_config:
-    backend: CUTLASS
+    backend: TRITON
   cuda_graph_config:
     enable_padding: true
     batch_sizes:

tests/integration/defs/disaggregated/test_configs/disagg_config_ctxtp2_gentp2_gptoss_eagle_trtllm.yaml

@@ -0,0 +1,67 @@
+model: gpt_oss/gpt-oss-120b
+hostname: localhost
+backend: pytorch
+context_servers:
+  num_instances: 1
+  tensor_parallel_size: 4
+  pipeline_parallel_size: 1
+  moe_expert_parallel_size: 4
+  enable_attention_dp: true
+  max_num_tokens: 16640
+  max_seq_len: 8232
+  max_batch_size: 128
+  trust_remote_code: true
+  enable_chunked_prefill: true
+  disable_overlap_scheduler: true
+  speculative_config: &eagle_spec
+    decoding_type: Eagle
+    max_draft_len: 3
+    eagle3_one_model: true
+    speculative_model: gpt_oss/gpt-oss-120b-Eagle3
+  kv_cache_config:
+    enable_block_reuse: false
+    free_gpu_memory_fraction: 0.8
+  moe_config:
+    backend: TRTLLM
+  cuda_graph_config: null
+  print_iter_log: true
+  cache_transceiver_config:
+    backend: DEFAULT
+    max_tokens_in_buffer: 16384
+generation_servers:
+  num_instances: 1
+  tensor_parallel_size: 4
+  pipeline_parallel_size: 1
+  moe_expert_parallel_size: 4
+  enable_attention_dp: true
+  max_num_tokens: 10240
+  max_seq_len: 10240
+  max_batch_size: 128
+  trust_remote_code: true
+  enable_chunked_prefill: true
+  disable_overlap_scheduler: true
+  speculative_config: *eagle_spec
+  kv_cache_config:
+    enable_block_reuse: false
+    free_gpu_memory_fraction: 0.8
+  moe_config:
+    backend: TRTLLM
+  cuda_graph_config:
+    enable_padding: true
+    batch_sizes:
+    - 1
+    - 2
+    - 4
+    - 8
+    - 16
+    - 32
+    - 64
+    - 128
+    - 256
+    - 512
+    - 768
+    - 1024
+  print_iter_log: true
+  cache_transceiver_config:
+    backend: DEFAULT
+    max_tokens_in_buffer: 16384

tests/integration/defs/disaggregated/test_configs/disagg_config_ctxtp2_gentp2_gptoss_tllm_cutlass.yaml → tests/integration/defs/disaggregated/test_configs/disagg_config_ctxtp2_gentp2_gptoss_triton.yaml

@@ -17,7 +17,7 @@ context_servers:
     free_gpu_memory_fraction: 0.8
   disable_overlap_scheduler: true
   moe_config:
-    backend: CUTLASS
+    backend: TRITON
   cuda_graph_config: null
   print_iter_log: true
   cache_transceiver_config:
@@ -39,7 +39,7 @@ generation_servers:
     free_gpu_memory_fraction: 0.8
   disable_overlap_scheduler: true
   moe_config:
-    backend: CUTLASS
+    backend: TRITON
   cuda_graph_config:
     enable_padding: true
     batch_sizes:

tests/integration/defs/disaggregated/test_disaggregated.py

@@ -298,12 +298,14 @@ def get_test_config(test_desc, example_dir, test_root):
         f"{test_configs_root}/disagg_config_ctxtp4_gentp4_deepseek_r1_v2_fp4_tllm.yaml",
         "deepseek_r1_v2_fp4_mtp_stress":
         f"{test_configs_root}/disagg_config_ctxtp4_gentp4_deepseek_r1_v2_fp4_tllm_mtp.yaml",
-        "gpt_oss_120b_stress":
+        "gpt_oss_120b_trtllm_stress":
         f"{test_configs_root}/disagg_config_ctxtp2_gentp2_gptoss_tllm.yaml",
-        "gpt_oss_120b_eagle_stress":
-        f"{test_configs_root}/disagg_config_ctxtp2_gentp2_gptoss_tllm_eagle.yaml",
-        "gpt_oss_120b_cutlass_stress":
-        f"{test_configs_root}/disagg_config_ctxtp2_gentp2_gptoss_tllm_cutlass.yaml",
+        "gpt_oss_120b_eagle_triton_stress":
+        f"{test_configs_root}/disagg_config_ctxtp2_gentp2_gptoss_eagle_triton.yaml",
+        "gpt_oss_120b_eagle_trtllm_stress":
+        f"{test_configs_root}/disagg_config_ctxtp2_gentp2_gptoss_eagle_trtllm.yaml",
+        "gpt_oss_120b_triton_stress":
+        f"{test_configs_root}/disagg_config_ctxtp2_gentp2_gptoss_triton.yaml",
         "qwen3_5_4b_fp8_stress":
         f"{test_configs_root}/disagg_config_ctxtp1_gentp1_qwen3_5_4b_fp8_tllm.yaml",
         "gpt_oss_120b_harmony":
@@ -2297,7 +2299,7 @@ def test_disaggregated_gpt_oss_120b_harmony(disaggregated_test_root,
                             cancellation_delay=0.5),
                  marks=(pytest.mark.skip_less_device(8), skip_pre_blackwell)),
     pytest.param(TestConfig(model_path='gpt_oss/gpt-oss-120b',
-                            test_desc='gpt_oss_120b_stress',
+                            test_desc='gpt_oss_120b_trtllm_stress',
                             request_count=60000,
                             accuracy_threshold=0.42,
                             cancellation_rate=10,
@@ -2306,22 +2308,34 @@ def test_disaggregated_gpt_oss_120b_harmony(disaggregated_test_root,
     pytest.param(
         TestConfig(
             model_path='gpt_oss/gpt-oss-120b',
-            test_desc='gpt_oss_120b_eagle_stress',
+            test_desc='gpt_oss_120b_eagle_triton_stress',
             request_count=60000,
             accuracy_threshold=0.42,
             speculative_model_path='gpt_oss/gpt-oss-120b-Eagle3',
             cancellation_rate=10,
             cancellation_delay=0.5,
         ),
-        marks=(pytest.mark.skip_less_device(8), skip_pre_hopper),
+        marks=(pytest.mark.skip_less_device(8), skip_no_hopper),
+    ),
+    pytest.param(
+        TestConfig(
+            model_path='gpt_oss/gpt-oss-120b',
+            test_desc='gpt_oss_120b_eagle_trtllm_stress',
+            request_count=60000,
+            accuracy_threshold=0.42,
+            speculative_model_path='gpt_oss/gpt-oss-120b-Eagle3',
+            cancellation_rate=10,
+            cancellation_delay=0.5,
+        ),
+        marks=(pytest.mark.skip_less_device(8), skip_pre_blackwell),
     ),
     pytest.param(TestConfig(model_path='gpt_oss/gpt-oss-120b',
-                            test_desc='gpt_oss_120b_cutlass_stress',
+                            test_desc='gpt_oss_120b_triton_stress',
                             request_count=60000,
                             accuracy_threshold=0.42,
                             cancellation_rate=10,
                             cancellation_delay=0.5),
-                 marks=(pytest.mark.skip_less_device(4), skip_pre_hopper)),
+                 marks=(pytest.mark.skip_less_device(4), skip_no_hopper)),
     pytest.param(TestConfig(model_path='Qwen3.5-4B-FP8',
                             test_desc='qwen3_5_4b_fp8_stress',
                             request_count=3000,

tests/integration/test_lists/qa/llm_function_core.txt

@@ -661,7 +661,6 @@ accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp4_tp2pp2[torc
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp4_tp2pp2[torch_compile=True-enable_gemm_allreduce_fusion=False]
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp4_tp2pp2[torch_compile=True-enable_gemm_allreduce_fusion=True]
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=False-torch_compile=False]
-accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=False-torch_compile=True]
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=True-torch_compile=False]
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=True-torch_compile=True]
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_tp4[torch_compile=False]

tests/integration/test_lists/qa/llm_function_rtx6k.txt

@@ -123,7 +123,6 @@ accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_nvfp4_tp4[torch
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=True-torch_compile=False]
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=True-torch_compile=True]
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=False-torch_compile=False]
-accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=False-torch_compile=True]
 accuracy/test_llm_api_pytorch.py::TestPhi4MiniInstruct::test_auto_dtype
 accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[latency_moe_cutlass-torch_compile=False]
 accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[latency_moe_cutlass-torch_compile=True]

tests/integration/test_lists/qa/llm_function_stress.txt

@@ -1,8 +1,9 @@
 disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-deepseek_r1_v2_fp4_stress]
 disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-deepseek_r1_v2_fp4_mtp_stress]
-disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_stress]
-disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_eagle_stress]
-disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_cutlass_stress]
+disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_trtllm_stress]
+disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_eagle_triton_stress]
+disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_eagle_trtllm_stress]
+disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_triton_stress]
 disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-qwen3_5_4b_fp8_stress]
 accuracy/test_llm_api_pytorch.py::TestDeepSeekR1LongBenchV2::test_fp8_8gpus
 accuracy/test_llm_api_pytorch.py::TestDeepSeekR1LongBenchV2::test_nvfp4_4gpus

tests/integration/test_lists/test-db/l0_l40s.yml

@@ -60,7 +60,6 @@ l0_l40s:
   - examples/test_llama.py::test_llm_llama_v2_lora_1gpu[chinese-llama-2-lora-13b-llama-v2-13b-hf-lora_fp16-base_fp16]
   - examples/test_llama.py::test_llm_llama_v3_dora_1gpu[commonsense-llama-v3-8b-dora-r32-llama-v3-8b-hf-base_fp16]
   - examples/test_llama.py::test_llm_llama_v1_1gpu_kv_cache_reuse_with_prompt_table[llama-7b]
-  - examples/test_phi.py::test_llm_phi_lora_1gpu[Phi-3-mini-4k-instruct-ru-lora-Phi-3-mini-4k-instruct-lora_fp16-base_fp16]
   - examples/test_nemotron_nas.py::test_nemotron_nas_summary_1gpu[DeciLM-7B]
   - llmapi/test_llm_examples.py::test_llmapi_quickstart
   - examples/test_visual_gen.py::test_visual_gen_quickstart

tests/integration/test_lists/waives.txt

@@ -116,7 +116,6 @@ accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_pard_sa SKIP (ht
 accuracy/test_llm_api_pytorch.py::TestLlama3_2_3B::test_auto_dtype SKIP (https://nvbugs/5520319)
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp4_tp2pp2[torch_compile=False-enable_gemm_allreduce_fusion=False] SKIP (https://nvbugs/6162122)
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp4_tp2pp2[torch_compile=False-enable_gemm_allreduce_fusion=True] SKIP (https://nvbugs/6162122)
-accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_eagle3_tp8[eagle3_one_model=False-torch_compile=True] SKIP (https://nvbugs/5775326)
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_fp8_tp4[torch_compile=True] SKIP (https://nvbugs/5821415)
 accuracy/test_llm_api_pytorch.py::TestLlama3_3_70BInstruct::test_nvfp4_tp4[torch_compile=True] SKIP (https://nvbugs/5821415)
 accuracy/test_llm_api_pytorch.py::TestMiniMaxM2::test_4gpus[attention_dp=False-cuda_graph=True-overlap_scheduler=True-tp_size=4-ep_size=4] SKIP (https://nvbugs/6159132)
@@ -203,7 +202,6 @@ examples/test_multimodal.py::test_llm_multimodal_general[video-neva-pp:1-tp:1-bf
 examples/test_nemotron.py::test_llm_nemotron_3_8b_1gpu[bfloat16-fp8] SKIP (https://nvbugs/4961624)
 examples/test_nemotron_nas.py::test_nemotron_nas_summary_1gpu[DeciLM-7B] SKIP (https://nvbugs/5444636)
 examples/test_nemotron_nas.py::test_nemotron_nas_summary_2gpu[DeciLM-7B] SKIP (https://nvbugs/5444636)
-examples/test_phi.py::test_llm_phi_lora_1gpu[Phi-3-mini-4k-instruct-ru-lora-Phi-3-mini-4k-instruct-lora_fp16-base_fp16] SKIP (https://nvbugs/5612313)
 examples/test_qwen2audio.py::test_llm_qwen2audio_single_gpu[qwen2_audio_7b_instruct] SKIP (https://nvbugs/5447530)
 examples/test_ray.py::test_ray_disaggregated_serving[tp2] SKIP (https://nvbugs/5612502)
 examples/test_recurrentgemma.py::test_llm_recurrentgemma_1gpu[use_cpp_session-recurrentgemma-2b-use_paged_cache-disable_quant-float16-enable_attn_plugin-enable_gemm_plugin] SKIP (https://nvbugs/5174573)
@@ -232,8 +230,8 @@ full:GH200/examples/test_multimodal.py::test_llm_multimodal_general[video-neva-p
 full:GH200/examples/test_nemotron.py::test_llm_nemotron_3_8b_1gpu[bfloat16-fp8] SKIP (arm is not supported)
 full:GH200/examples/test_qwen2audio.py::test_llm_qwen2audio_single_gpu[qwen2_audio_7b_instruct] SKIP (arm is not supported)
 full:GH200/unittest/trt/model_api/test_model_quantization.py SKIP (https://nvbugs/4979955)
-full:H100/disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_cutlass_stress] SKIP (https://nvbugs/6222480)
-full:H100/disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_eagle_stress] SKIP (https://nvbugs/6222480)
+full:H100/disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_eagle_triton_stress] SKIP (https://nvbugs/6250439)
+full:H100/disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_triton_stress] SKIP (https://nvbugs/6222480)
 full:H100_PCIe/unittest/llmapi/test_llm_pytorch.py::test_llama_7b_multi_lora_evict_and_reload_lora_gpu_cache SKIP (https://nvbugs/5682551)
 full:H20/accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_chunked_prefill[triton-auto] SKIP (https://nvbugs/6026676)
 full:RTX/accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype SKIP (https://nvbugs/5569696)