[None][feat] adding support for disaggregated multi-instance tests (NVIDIA#6674)

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

Author: raayandhar

tests/integration/defs/accuracy/test_disaggregated_serving.py

@@ -847,6 +847,43 @@ def test_auto_dtype(self, overlap_scheduler):
             task = MMLU(self.MODEL_NAME)
             task.evaluate(llm)
 
+    def test_chunked_prefill(self):
+        ctx_server_config = {
+            "disable_overlap_scheduler": True,
+            "cuda_graph_config": None,
+            "cache_transceiver_config": {
+                "backend": "DEFAULT"
+            },
+            "enable_chunked_prefill": True,
+            "max_num_tokens": 256,
+        }
+        gen_server_config = {
+            "cuda_graph_config": None,
+            "cache_transceiver_config": {
+                "backend": "DEFAULT"
+            }
+        }
+        disaggregated_server_config = {
+            "hostname": "localhost",
+            "port": 8000,
+            "backend": "pytorch",
+            "context_servers": {
+                "num_instances": 1,
+                "urls": ["localhost:8001"]
+            },
+            "generation_servers": {
+                "num_instances": 1,
+                "urls": ["localhost:8002"]
+            }
+        }
+        with launch_disaggregated_llm(disaggregated_server_config,
+                                      ctx_server_config, gen_server_config,
+                                      self.MODEL_PATH) as llm:
+            task = GSM8K(self.MODEL_NAME)
+            task.evaluate(llm)
+            task = MMLU(self.MODEL_NAME)
+            task.evaluate(llm)
+
 
 @skip_pre_blackwell
 @pytest.mark.timeout(3600)
@@ -868,4 +905,3 @@ def test_mixed_ctx_gen_model(self, ctx_pp, gen_tp):
                                  test_sets=[GSM8K, MMLU],
                                  ctx_model=ctx_model,
                                  gen_model=gen_model)
-

tests/integration/test_lists/qa/llm_function_full.txt

@@ -572,6 +572,8 @@ accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_ctx_pp_gen
 accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_ctx_pp_gen_tp_asymmetric[MMLU-gen_tp=1-ctx_pp=4]
 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_ctx_pp_gen_tp_asymmetric[MMLU-gen_tp=2-ctx_pp=4]
+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_disaggregated_serving.py::TestLlama4ScoutInstruct::test_auto_dtype[False]
 accuracy/test_disaggregated_serving.py::TestLlama4ScoutInstruct::test_auto_dtype[True]
 accuracy/test_disaggregated_serving.py::TestQwen3_8B::test_auto_dtype[False]

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

@@ -54,6 +54,8 @@ l0_dgx_h100:
   - accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_ctx_pp_gen_tp_asymmetric[GSM8K-gen_tp=2-ctx_pp=2]
   - accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_ctx_pp_gen_tp_asymmetric[MMLU-gen_tp=1-ctx_pp=2]
   - 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_disaggregated_serving.py::TestQwen3_8B::test_nixl_backend
   - accuracy/test_disaggregated_serving.py::TestDeepSeekV3Lite::test_nixl_backend
   - test_e2e.py::test_ptp_quickstart_advanced_bs1