From 6b9b73ee27b365ddffc54b9c2813f553df6770f6 Mon Sep 17 00:00:00 2001 From: ruodil <200874449+ruodil@users.noreply.github.com> Date: Tue, 28 Oct 2025 15:25:52 +0800 Subject: [PATCH] [https://nvbugs/5564465][test] ensure deepseek_v3_lite isl + osl < max_seq_len (#8565) Signed-off-by: Ruodi Lu Co-authored-by: Ruodi Lu --- tests/integration/test_lists/qa/llm_perf_cluster.yml | 2 +- tests/integration/test_lists/qa/llm_perf_cluster_nim.yml | 2 +- tests/integration/test_lists/qa/llm_perf_core.yml | 2 +- tests/integration/test_lists/qa/llm_perf_sanity.yml | 2 +- 4 files changed, 4 insertions(+), 4 deletions(-) diff --git a/tests/integration/test_lists/qa/llm_perf_cluster.yml b/tests/integration/test_lists/qa/llm_perf_cluster.yml index f5a895cb47..796b4d8ef3 100644 --- a/tests/integration/test_lists/qa/llm_perf_cluster.yml +++ b/tests/integration/test_lists/qa/llm_perf_cluster.yml @@ -17,7 +17,7 @@ llm_perf_cluster: - perf/test_perf.py::test_perf[mistral_7b_v0.1-bench-float16-input_output_len:500,2000-quant:fp8] - perf/test_perf.py::test_perf[deepseek_v3_lite_nvfp4-bench-pytorch-streaming-float4-maxbs:2048-maxnt:8192-input_output_len:256,256-reqs:200] # for chunked prefill cases - - perf/test_perf.py::test_perf[deepseek_v3_lite_nvfp4-bench-pytorch-float4-maxbs:512-maxnt:2048-kv_frac:0.85-input_output_len:5000,500-reqs:200] + - perf/test_perf.py::test_perf[deepseek_v3_lite_nvfp4-bench-pytorch-float4-maxbs:512-maxnt:2048-kv_frac:0.85-input_output_len:3000,500-reqs:200] # Phi-4-multimodal-instruct - perf/test_perf.py::test_perf[phi_4_multimodal_instruct-bench-pytorch-bfloat16-input_output_len:500,2000-con:250] - perf/test_perf.py::test_perf[phi_4_multimodal_instruct-bench-pytorch-bfloat16-input_output_len:1000,1000-con:250] diff --git a/tests/integration/test_lists/qa/llm_perf_cluster_nim.yml b/tests/integration/test_lists/qa/llm_perf_cluster_nim.yml index e56252fd7e..b938600890 100644 --- a/tests/integration/test_lists/qa/llm_perf_cluster_nim.yml +++ b/tests/integration/test_lists/qa/llm_perf_cluster_nim.yml @@ -17,7 +17,7 @@ llm_perf_cluster_nim: - perf/test_perf.py::test_perf[mistral_7b_v0.1-bench-float16-input_output_len:500,2000-quant:fp8] - perf/test_perf.py::test_perf[deepseek_v3_lite_nvfp4-bench-pytorch-streaming-float4-maxbs:2048-maxnt:8192-input_output_len:256,256-reqs:200] # for chunked prefill cases - - perf/test_perf.py::test_perf[deepseek_v3_lite_nvfp4-bench-pytorch-float4-maxbs:512-maxnt:2048-kv_frac:0.85-input_output_len:5000,500-reqs:200] + - perf/test_perf.py::test_perf[deepseek_v3_lite_nvfp4-bench-pytorch-float4-maxbs:512-maxnt:2048-kv_frac:0.85-input_output_len:3000,500-reqs:200] # Phi-4-multimodal-instruct - perf/test_perf.py::test_perf[phi_4_multimodal_instruct-bench-pytorch-bfloat16-input_output_len:500,2000-con:250] - perf/test_perf.py::test_perf[phi_4_multimodal_instruct-bench-pytorch-bfloat16-input_output_len:1000,1000-con:250] diff --git a/tests/integration/test_lists/qa/llm_perf_core.yml b/tests/integration/test_lists/qa/llm_perf_core.yml index e888c70e2b..c04b6d87c5 100644 --- a/tests/integration/test_lists/qa/llm_perf_core.yml +++ b/tests/integration/test_lists/qa/llm_perf_core.yml @@ -80,7 +80,7 @@ llm_perf_core: - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-float8-input_output_len:2000,500] - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-streaming-float8-input_output_len:2000,500] - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-float8-input_output_len:500,2000] - - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-float8-maxbs:512-maxnt:2048-kv_frac:0.85-input_output_len:5000,500-reqs:200] + - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-float8-maxbs:512-maxnt:2048-kv_frac:0.85-input_output_len:3000,500-reqs:200] - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-streaming-float8-input_output_len:128,128] - perf/test_perf.py::test_perf[mistral_small_v3.1_24b-bench-pytorch-bfloat16-maxbs:1-maxnt:20000-input_output_len:20000,2000-reqs:8-con:1] TIMEOUT(120) - perf/test_perf.py::test_perf[mistral_small_v3.1_24b-bench-pytorch-bfloat16-maxbs:4096-maxnt:20000-input_output_len:20000,2000-reqs:500-con:200] TIMEOUT(120) diff --git a/tests/integration/test_lists/qa/llm_perf_sanity.yml b/tests/integration/test_lists/qa/llm_perf_sanity.yml index 02270c421f..3abd7e66f2 100644 --- a/tests/integration/test_lists/qa/llm_perf_sanity.yml +++ b/tests/integration/test_lists/qa/llm_perf_sanity.yml @@ -166,7 +166,7 @@ llm_perf_sanity: - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-streaming-float8-input_output_len:128,128] - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-streaming-float8-input_output_len:2000,500] # for chunked prefill cases - - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-float8-maxbs:512-maxnt:2048-kv_frac:0.85-input_output_len:5000,500-reqs:200] + - perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-float8-maxbs:512-maxnt:2048-kv_frac:0.85-input_output_len:3000,500-reqs:200] - perf/test_perf.py::test_perf[llama_v3.1_405b_instruct_fp8-bench-pytorch-float8-input_output_len:128,128-tp:8-gpus:8] - perf/test_perf.py::test_perf[llama_v4_maverick_17b_128e_instruct_fp8-bench-pytorch-float8-input_output_len:512,32-ep:8-tp:8-gpus:8] - perf/test_perf.py::test_perf[llama_v4_maverick_17b_128e_instruct_fp8-bench-pytorch-float8-maxbs:1024-maxnt:20000-kv_frac:0.6-input_output_len:20000,2000-reqs:1000-ep:8-tp:8-gpus:8]