add tests in different gpus

Signed-off-by: jiant <[email protected]>

Author: JadoTu

tests/integration/defs/accuracy/references/gsm8k.yaml

@@ -139,10 +139,10 @@ Qwen3/Qwen3-235B-A22B:
 Qwen3/Qwen3-Next-80B-A3B-Thinking:
   - accuracy: 81.577
 Qwen3/Qwen3-Next-80B-A3B-Instruct:
-  - accuracy: 84.42
+  - accuracy: 83.36
   - quant_algo: NVFP4
     kv_cache_quant_algo: FP8
-    accuracy: 84.32
+    accuracy: 80.84
 moonshotai/Kimi-K2-Instruct:
   - quant_algo: FP8_BLOCK_SCALES
     accuracy: 94.84

tests/integration/defs/accuracy/references/mmlu.yaml

@@ -243,10 +243,10 @@ Qwen3/Qwen3-235B-A22B:
 Qwen3/Qwen3-Next-80B-A3B-Thinking:
   - accuracy: 86
 Qwen3/Qwen3-Next-80B-A3B-Instruct:
-  - accuracy: 85.58
+  - accuracy: 86.03
   - quant_algo: NVFP4
     kv_cache_quant_algo: FP8
-    accuracy: 85
+    accuracy: 85.08
 moonshotai/Kimi-K2-Instruct:
   - quant_algo: FP8_BLOCK_SCALES
     accuracy: 87.65

tests/integration/defs/accuracy/test_llm_api_pytorch.py

@@ -4157,7 +4157,6 @@ class TestQwen3NextInstruct(LlmapiAccuracyTestHarness):
     def test_bf16_4gpu(self, tp_size, pp_size, ep_size, cuda_graph,
                        overlap_scheduler):
         model_path = f"{self.MODEL_PATH}/Qwen3-Next-80B-A3B-Instruct"
-        model_path = "Qwen/Qwen3-Next-80B-A3B-Instruct"
         kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.6,
                                         enable_block_reuse=False)
         pytorch_config = dict(disable_overlap_scheduler=not overlap_scheduler,
@@ -4189,7 +4188,6 @@ def test_bf16_4gpu(self, tp_size, pp_size, ep_size, cuda_graph,
     def test_nvfp4(self, moe_backend, tp_size, pp_size, ep_size, cuda_graph,
                    overlap_scheduler):
         model_path = f"{self.MODEL_PATH}/qwen3-next-80b-instruct-nvfp4-ptq-fp8kv"
-        model_path = "/home/scratch.didow_sw_1/models/qwen3-next-80b-instruct-nvfp4-ptq-fp8kv"
 
         kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.6,
                                         enable_block_reuse=False)

tests/integration/test_lists/test-db/l0_b200.yml

@@ -54,6 +54,11 @@ l0_b200:
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a8_mxfp4[mxfp8-latency-TRTLLM]
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a8_mxfp4[mxfp8-latency-CUTLASS]
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_w4a16_mxfp4[latency-TRTLLM]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[no_cuda_graph_overlap-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-trtllm]
   - disaggregated/test_workers.py::test_workers_kv_cache_aware_router_eviction[TinyLlama-1.1B-Chat-v1.0] # nvbugs 5300551
   - 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[Llama3.1-8B-FP8-llama-3.1-model/Llama-3.1-8B-Instruct-FP8]

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

@@ -42,6 +42,12 @@ 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::TestQwen3_30B_A3B::test_nvfp4[dep4_latency_moe_cutlass-torch_compile=True] ISOLATION
   - accuracy/test_llm_api_pytorch.py::TestQwen3NextThinking::test_auto_dtype[tp4ep4]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_bf16_4gpu[tp4ep4_cudagraph_overlap]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[no_cuda_graph_overlap-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-trtllm]
   - 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-auto]
   - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[ep4-cutlass-auto]

tests/integration/test_lists/test-db/l0_dgx_b300.yml

@@ -58,6 +58,11 @@ l0_dgx_b300:
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[dep4_latency_moe_trtllm-torch_compile=False]
   - accuracy/test_llm_api_pytorch.py::TestQwen3_30B_A3B::test_nvfp4[dep4_latency_moe_cutlass-torch_compile=True]
   - accuracy/test_disaggregated_serving.py::TestQwen3_30B_A3B::test_mixed_ctx_gen_model[ctxpp2gentp2]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[no_cuda_graph_overlap-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-trtllm]
   - 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-cutlass-fp8]
   - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-cutlass-auto]

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

@@ -93,6 +93,7 @@ l0_dgx_h100:
   - accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_ctx_pp_gen_tp_asymmetric[MMLU-gen_tp=2-ctx_pp=2]
   - accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_multi_instance[GSM8K]
   - accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_multi_instance[MMLU]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_bf16_4gpu[tp4ep4_cudagraph_overlap]
   - disaggregated/test_auto_scaling.py::test_service_discovery[etcd-round_robin]
   - disaggregated/test_auto_scaling.py::test_service_discovery[etcd-load_balancing]
   - disaggregated/test_auto_scaling.py::test_worker_restart[etcd-round_robin]

tests/integration/test_lists/test-db/l0_dgx_h200.yml

@@ -33,6 +33,7 @@ l0_dgx_h200:
   - accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_ctx_pp_gen_tp_asymmetric[MMLU-gen_tp=2-ctx_pp=4]
   - accuracy/test_disaggregated_serving.py::TestGPTOSS::test_auto_dtype[False]
   - accuracy/test_disaggregated_serving.py::TestGPTOSS::test_auto_dtype[True]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_bf16_4gpu[tp4ep4_cudagraph_overlap]
   - disaggregated/test_disaggregated.py::test_disaggregated_ctxtp2pp2_gentp2pp2[TinyLlama-1.1B-Chat-v1.0]
   - disaggregated/test_disaggregated.py::test_disaggregated_ctxpp4_genpp4[TinyLlama-1.1B-Chat-v1.0]
   - unittest/llmapi/test_llm_pytorch.py::test_nemotron_nas_lora

tests/integration/test_lists/test-db/l0_gb200_multi_gpus.yml

@@ -43,6 +43,11 @@ l0_gb200_multi_gpus:
   - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[tp4-trtllm-auto]
   - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[dp4-cutlass-auto]
   - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus_online_eplb[fp8]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[no_cuda_graph_overlap-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-trtllm]
 - condition:
     ranges:
       system_gpu_count:

tests/integration/test_lists/test-db/l0_rtx_pro_6000.yml

@@ -43,6 +43,8 @@ l0_rtx_pro_6000:
   - accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu[True-True-cutlass-auto]
   - accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_fp4
   - accuracy/test_llm_api_pytorch.py::TestPhi4MM::test_fp8
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp1-cutlass]
+  - accuracy/test_llm_api_pytorch.py::TestQwen3NextInstruct::test_nvfp4[tp4ep4-trtllm]
 
 - condition:
     ranges: