From 5c9d1e4b5c72628a5f17940fc64db9fa7ea0e199 Mon Sep 17 00:00:00 2001 From: raayandhar Date: Wed, 30 Jul 2025 09:26:06 -0700 Subject: [PATCH 1/6] adding the fix for the disagg pp4 test Signed-off-by: raayandhar --- tests/integration/defs/disaggregated/test_disaggregated.py | 5 ++--- tests/integration/test_lists/test-db/l0_dgx_h200.yml | 2 +- 2 files changed, 3 insertions(+), 4 deletions(-) diff --git a/tests/integration/defs/disaggregated/test_disaggregated.py b/tests/integration/defs/disaggregated/test_disaggregated.py index 24fdfcfad9e1..c8ffb394f463 100644 --- a/tests/integration/defs/disaggregated/test_disaggregated.py +++ b/tests/integration/defs/disaggregated/test_disaggregated.py @@ -632,15 +632,14 @@ def test_disaggregated_ctxtp2pp2_gentp2pp2(disaggregated_test_root, llm_venv, @pytest.mark.skip_less_device(8) -@pytest.mark.parametrize("llama_model_root", ['TinyLlama-1.1B-Chat-v1.0'], +@pytest.mark.parametrize("llama_model_root", ['Llama-3.1-8B-Instruct'], indirect=True) def test_disaggregated_ctxpp4_genpp4(disaggregated_test_root, llm_venv, disaggregated_example_root, llama_model_root): - pytest.skip(f"8 GPU test times out currently, skipping") src_dst_dict = { llama_model_root: - f"{llm_venv.get_working_directory()}/TinyLlama/TinyLlama-1.1B-Chat-v1.0", + f"{llm_venv.get_working_directory()}/meta-llama/Llama-3.1-8B-Instruct", } for src, dst in src_dst_dict.items(): if not os.path.islink(dst): diff --git a/tests/integration/test_lists/test-db/l0_dgx_h200.yml b/tests/integration/test_lists/test-db/l0_dgx_h200.yml index 33542dd8d751..48e052c6ce7d 100644 --- a/tests/integration/test_lists/test-db/l0_dgx_h200.yml +++ b/tests/integration/test_lists/test-db/l0_dgx_h200.yml @@ -30,7 +30,7 @@ l0_dgx_h200: - 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=4] - 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] + - disaggregated/test_disaggregated.py::test_disaggregated_ctxpp4_genpp4[Llama-3.1-8B-Instruct] - unittest/_torch/multi_gpu_modeling/test_llama4.py::test_llama4[pp1-ep1-disable_adp-enable_graph-tp8-trtllm-scout] - unittest/_torch/multi_gpu_modeling/test_llama4.py::test_llama4[pp1-ep4-enable_adp-enable_graph-tp8-trtllm-scout] - unittest/llmapi/test_llm_pytorch.py::test_nemotron_nas_lora From e3faf3ac7bfe2ab85e7ba90d7869e20476aa07b7 Mon Sep 17 00:00:00 2001 From: raayandhar Date: Wed, 30 Jul 2025 20:58:58 +0000 Subject: [PATCH 2/6] another test was being skipped that shouldn't have been Signed-off-by: raayandhar --- tests/integration/defs/disaggregated/test_disaggregated.py | 1 - 1 file changed, 1 deletion(-) diff --git a/tests/integration/defs/disaggregated/test_disaggregated.py b/tests/integration/defs/disaggregated/test_disaggregated.py index c8ffb394f463..df3006c23317 100644 --- a/tests/integration/defs/disaggregated/test_disaggregated.py +++ b/tests/integration/defs/disaggregated/test_disaggregated.py @@ -616,7 +616,6 @@ def test_disaggregated_ctxpp2_gentp2(disaggregated_test_root, llm_venv, def test_disaggregated_ctxtp2pp2_gentp2pp2(disaggregated_test_root, llm_venv, disaggregated_example_root, llama_model_root): - pytest.skip(f"8 GPU test times out currently, skipping") src_dst_dict = { llama_model_root: f"{llm_venv.get_working_directory()}/TinyLlama/TinyLlama-1.1B-Chat-v1.0", From e589ff7082473996eda3b45bc3ef1a31902aad14 Mon Sep 17 00:00:00 2001 From: raayandhar Date: Wed, 30 Jul 2025 22:54:37 +0000 Subject: [PATCH 3/6] choose a model that works with conftest.py Signed-off-by: raayandhar --- tests/integration/defs/disaggregated/test_disaggregated.py | 5 ++--- tests/integration/test_lists/test-db/l0_dgx_h200.yml | 2 +- 2 files changed, 3 insertions(+), 4 deletions(-) diff --git a/tests/integration/defs/disaggregated/test_disaggregated.py b/tests/integration/defs/disaggregated/test_disaggregated.py index df3006c23317..fbdf21eea2ed 100644 --- a/tests/integration/defs/disaggregated/test_disaggregated.py +++ b/tests/integration/defs/disaggregated/test_disaggregated.py @@ -631,14 +631,13 @@ def test_disaggregated_ctxtp2pp2_gentp2pp2(disaggregated_test_root, llm_venv, @pytest.mark.skip_less_device(8) -@pytest.mark.parametrize("llama_model_root", ['Llama-3.1-8B-Instruct'], - indirect=True) +@pytest.mark.parametrize("llama_model_root", ['llama-3.2-1b'], indirect=True) def test_disaggregated_ctxpp4_genpp4(disaggregated_test_root, llm_venv, disaggregated_example_root, llama_model_root): src_dst_dict = { llama_model_root: - f"{llm_venv.get_working_directory()}/meta-llama/Llama-3.1-8B-Instruct", + f"{llm_venv.get_working_directory()}/llama-3.2-models/llama-3.2-1b", } for src, dst in src_dst_dict.items(): if not os.path.islink(dst): diff --git a/tests/integration/test_lists/test-db/l0_dgx_h200.yml b/tests/integration/test_lists/test-db/l0_dgx_h200.yml index 48e052c6ce7d..a8331ac0b9d5 100644 --- a/tests/integration/test_lists/test-db/l0_dgx_h200.yml +++ b/tests/integration/test_lists/test-db/l0_dgx_h200.yml @@ -30,7 +30,7 @@ l0_dgx_h200: - 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=4] - disaggregated/test_disaggregated.py::test_disaggregated_ctxtp2pp2_gentp2pp2[TinyLlama-1.1B-Chat-v1.0] - - disaggregated/test_disaggregated.py::test_disaggregated_ctxpp4_genpp4[Llama-3.1-8B-Instruct] + - disaggregated/test_disaggregated.py::test_disaggregated_ctxpp4_genpp4[llama-3.2-1b] - unittest/_torch/multi_gpu_modeling/test_llama4.py::test_llama4[pp1-ep1-disable_adp-enable_graph-tp8-trtllm-scout] - unittest/_torch/multi_gpu_modeling/test_llama4.py::test_llama4[pp1-ep4-enable_adp-enable_graph-tp8-trtllm-scout] - unittest/llmapi/test_llm_pytorch.py::test_nemotron_nas_lora From 3ebdf5bcc96229880821d70f5775a8b09531245a Mon Sep 17 00:00:00 2001 From: raayandhar Date: Thu, 31 Jul 2025 16:13:10 +0000 Subject: [PATCH 4/6] llama 3.2 1b hangs, switching to 3.1 8b works Signed-off-by: raayandhar --- tests/integration/defs/disaggregated/test_disaggregated.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tests/integration/defs/disaggregated/test_disaggregated.py b/tests/integration/defs/disaggregated/test_disaggregated.py index fbdf21eea2ed..00cadc903391 100644 --- a/tests/integration/defs/disaggregated/test_disaggregated.py +++ b/tests/integration/defs/disaggregated/test_disaggregated.py @@ -631,13 +631,13 @@ def test_disaggregated_ctxtp2pp2_gentp2pp2(disaggregated_test_root, llm_venv, @pytest.mark.skip_less_device(8) -@pytest.mark.parametrize("llama_model_root", ['llama-3.2-1b'], indirect=True) +@pytest.mark.parametrize("llama_model_root", ['llama-3.1-8b'], indirect=True) def test_disaggregated_ctxpp4_genpp4(disaggregated_test_root, llm_venv, disaggregated_example_root, llama_model_root): src_dst_dict = { llama_model_root: - f"{llm_venv.get_working_directory()}/llama-3.2-models/llama-3.2-1b", + f"{llm_venv.get_working_directory()}/llama-3.1-models/Meta-Llama-3.1-8B", } for src, dst in src_dst_dict.items(): if not os.path.islink(dst): From 52728c3317d5026226c907c1524689aa4b91c43c Mon Sep 17 00:00:00 2001 From: Raayan Dhar <58057652+raayandhar@users.noreply.github.com> Date: Thu, 31 Jul 2025 14:07:58 -0700 Subject: [PATCH 5/6] Update l0_dgx_h200.yml Signed-off-by: Raayan Dhar <58057652+raayandhar@users.noreply.github.com> --- tests/integration/test_lists/test-db/l0_dgx_h200.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/integration/test_lists/test-db/l0_dgx_h200.yml b/tests/integration/test_lists/test-db/l0_dgx_h200.yml index a8331ac0b9d5..73af92d1ddf8 100644 --- a/tests/integration/test_lists/test-db/l0_dgx_h200.yml +++ b/tests/integration/test_lists/test-db/l0_dgx_h200.yml @@ -30,7 +30,7 @@ l0_dgx_h200: - 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=4] - disaggregated/test_disaggregated.py::test_disaggregated_ctxtp2pp2_gentp2pp2[TinyLlama-1.1B-Chat-v1.0] - - disaggregated/test_disaggregated.py::test_disaggregated_ctxpp4_genpp4[llama-3.2-1b] + - disaggregated/test_disaggregated.py::test_disaggregated_ctxpp4_genpp4[llama-3.1-8b] - unittest/_torch/multi_gpu_modeling/test_llama4.py::test_llama4[pp1-ep1-disable_adp-enable_graph-tp8-trtllm-scout] - unittest/_torch/multi_gpu_modeling/test_llama4.py::test_llama4[pp1-ep4-enable_adp-enable_graph-tp8-trtllm-scout] - unittest/llmapi/test_llm_pytorch.py::test_nemotron_nas_lora From 6066228b6f93ae28f45c366ac6cc73ee6d322528 Mon Sep 17 00:00:00 2001 From: raayandhar Date: Tue, 5 Aug 2025 11:10:47 -0700 Subject: [PATCH 6/6] fix the test list check that was failing Signed-off-by: raayandhar --- tests/integration/test_lists/qa/llm_function_full.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/integration/test_lists/qa/llm_function_full.txt b/tests/integration/test_lists/qa/llm_function_full.txt index ddd12c41688a..28e6196b292e 100644 --- a/tests/integration/test_lists/qa/llm_function_full.txt +++ b/tests/integration/test_lists/qa/llm_function_full.txt @@ -662,7 +662,7 @@ disaggregated/test_disaggregated_single_gpu.py::test_disaggregated_simple_qwen3[ disaggregated/test_disaggregated_single_gpu.py::test_disaggregated_simple_qwen3[True-False-Qwen3-8B-FP8] disaggregated/test_disaggregated_single_gpu.py::test_disaggregated_simple_qwen3[True-True-Qwen3-8B-FP8] 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] +disaggregated/test_disaggregated.py::test_disaggregated_ctxpp4_genpp4[llama-3.1-8b] disaggregated/test_workers.py::test_workers_conditional_disaggregation[TinyLlama-1.1B-Chat-v1.0] disaggregated/test_workers.py::test_workers_kv_cache_events[TinyLlama-1.1B-Chat-v1.0] disaggregated/test_workers.py::test_workers_kv_cache_aware_router[TinyLlama-1.1B-Chat-v1.0]