Skip to content

Tiny enhance test suites sanity check #13589

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
hnyls2002 merged 4 commits into from
Nov 19, 2025
Merged

Tiny enhance test suites sanity check #13589

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
56e9256
fix
hnyls2002 Nov 19, 2025
5750596
fix
hnyls2002 Nov 19, 2025
a1a6868
add more check
hnyls2002 Nov 19, 2025
111e07a
fix
hnyls2002 Nov 19, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
89 changes: 25 additions & 64 deletions test/srt/run_suite.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -230,6 +230,7 @@
"nightly-8-gpu-h20": [],
"__not_in_ci__": [
TestFile("ascend/test_ascend_w8a8_quantization.py"),
TestFile("ascend/test_mindspore_models.py"),
TestFile("cpu/test_comm.py"),
TestFile("debug_utils/test_log_parser.py", 5),
TestFile("test_deepseek_v3_cutedsl_4gpu.py"),
Expand All @@ -242,8 +243,6 @@
TestFile("ep/test_moe_deepep.py"),
TestFile("ep/test_moe_deepep_eval_accuracy_large.py"),
TestFile("hicache/test_disaggregation_hicache.py"),
TestFile("hicache/test_hicache_storage_benchmark.py"),
TestFile("hicache/test_hicache_storage_e2e.py"),
TestFile("layers/attention/nsa/test_act_quant_triton.py"),
TestFile("layers/moe/test_moe_runners.py"),
TestFile("lora/test_chunked_sgmv_backend.py"),
Expand All @@ -258,120 +257,61 @@
TestFile("models/test_llama4_models.py"),
TestFile("models/test_mtp_models.py"),
TestFile("models/test_unsloth_models.py"),
TestFile("openai/test_server.py"),
TestFile("openai_server/features/test_cache_report.py"),
TestFile("openai_server/features/test_continuous_usage_stats.py"),
TestFile("openai_server/features/test_structural_tag.py"),
TestFile("quant/test_fp8_kvcache.py"),
TestFile("rl/test_verl_engine_2_gpu.py"),
TestFile("rl/test_verl_engine_4_gpu.py"),
TestFile("test_ascend_attention_backend.py"),
TestFile("test_ascend_mla_backend.py"),
TestFile("test_ascend_mla_w8a8int8.py"),
TestFile("test_ascend_tp1_bf16.py"),
TestFile("test_ascend_tp2_bf16.py"),
TestFile("test_ascend_w8a8_quantization.py"),
TestFile("test_async_dynamic_batch_tokenizer.py"),
TestFile("test_async_mm_data_processor.py"),
TestFile("test_awq.py"),
TestFile("test_awq_dequant.py"),
TestFile("test_bench_one_batch.py"),
TestFile("test_bench_serving.py"),
TestFile("test_block_int8.py"),
TestFile("test_cache_report.py"),
TestFile("test_config_integration.py"),
TestFile("test_cpp_radix_cache.py"),
TestFile("test_cpu_graph.py"),
TestFile("test_custom_allreduce.py"),
TestFile("test_cutedsl_flashinfer_8gpu.py"),
TestFile("test_deepep_internode.py"),
TestFile("test_deepep_intranode.py"),
TestFile("test_deepep_large.py"),
TestFile("test_deepep_low_latency.py"),
TestFile("test_deepep_small.py"),
TestFile("test_deepseek_chat_templates.py"),
TestFile("test_disaggregation.py"),
TestFile("test_double_sparsity.py"),
TestFile("test_eagle_infer_beta_dp_attention.py"),
TestFile("test_embedding_openai_server.py"),
TestFile("test_enable_thinking.py"),
TestFile("test_eplb.py"),
TestFile("test_eval_accuracy_large.py"),
TestFile("test_expert_distribution.py"),
TestFile("test_expert_location_updater.py"),
TestFile("test_fim_completion.py"),
TestFile("test_forward_split_prefill.py"),
TestFile("test_fp8_kernel.py"),
TestFile("test_fp8_kvcache.py"),
TestFile("test_full_deepseek_v3.py"),
TestFile("test_get_weights_by_name.py"),
TestFile("test_gpt_oss_common.py"),
TestFile("test_health_check.py"),
TestFile("test_hicache_storage.py"),
TestFile("test_hicache_variants.py"),
TestFile("test_hybrid_dp_ep_tp_mtp.py"),
TestFile("test_int4_kernel.py"),
TestFile("test_int8_kernel.py"),
TestFile("test_intel_amx_attention_backend.py"),
TestFile("test_constrained_decoding.py"),
TestFile("test_json_mode.py"),
TestFile("test_kv_events.py"),
TestFile("test_large_max_new_tokens.py"),
TestFile("test_logprobs.py"),
TestFile("test_lookahead_speculative_decoding.py"),
TestFile("test_matched_stop.py"),
TestFile("test_mla_tp.py"),
TestFile("test_modelopt.py"),
TestFile("test_modelopt_fp8kvcache.py"),
TestFile("test_models_from_modelscope.py"),
TestFile("test_moe_deepep.py"),
TestFile("test_moe_deepep_eval_accuracy_large.py"),
TestFile("test_moe_ep.py"),
TestFile("test_moe_eval_accuracy_large.py"),
TestFile("test_mscclpp.py"),
TestFile("nightly/test_deepseek_v31_perf.py"),
TestFile("nightly/test_deepseek_v32_perf.py"),
TestFile("nightly/test_gpt_oss_4gpu_perf.py"),
TestFile("nightly/test_gsm8k_eval_amd.py"),
TestFile("nightly/test_text_models_gsm8k_eval.py"),
TestFile("nightly/test_text_models_perf.py"),
TestFile("nightly/test_vlms_mmmu_eval.py"),
TestFile("nightly/test_vlms_perf.py"),
TestFile("test_openai_adapter.py"),
TestFile("test_openai_function_calling.py"),
TestFile("test_openai_server.py"),
TestFile("test_openai_server_hidden_states.py"),
TestFile("test_piecewise_cuda_graph.py"),
TestFile("test_quick_allreduce.py"),
TestFile("test_reasoning_content.py"),
TestFile("test_request_length_validation.py"),
TestFile("test_sagemaker_server.py"),
TestFile("test_schedule_policy.py"),
TestFile("test_session_control.py"),
TestFile("test_srt_engine_with_quant_args.py"),
TestFile("test_tokenizer_batch_encode.py"),
TestFile("test_tokenizer_manager.py"),
TestFile("test_tool_choice.py"),
TestFile("test_torch_flex_attention_backend.py"),
TestFile("test_torch_tp.py"),
TestFile("test_tracing.py"),
TestFile("test_triton_attention_rocm_mla.py"),
TestFile("test_triton_fused_moe.py"),
TestFile("test_triton_moe_wna16.py"),
TestFile("test_two_batch_overlap.py"),
TestFile("test_update_weights_from_disk.py"),
TestFile("test_update_weights_from_distributed.py"),
TestFile("test_update_weights_from_tensor.py"),
TestFile("test_verl_engine_2_gpu.py"),
TestFile("test_verl_engine_4_gpu.py"),
TestFile("test_verl_engine_server.py"),
TestFile("test_vertex_endpoint.py"),
# TestFile("test_vision_openai_server_a.py"), # TODO: Fix timeout
TestFile("test_vision_openai_server_b.py"),
TestFile("test_vision_openai_server_common.py"),
TestFile("test_vlm_accuracy.py"),
TestFile("test_w4a8.py"),
TestFile("test_w8a8_quantization.py"),
TestFile("test_wave_attention_backend.py"),
TestFile("test_weight_version.py"),
TestFile("test_deepseek_v32_cp_single_node.py", 275),
Expand Down Expand Up @@ -537,9 +477,6 @@
TestFile("ascend/test_ascend_deepep.py", 400),
TestFile("ascend/test_ascend_deepseek_mtp.py", 400),
],
"__not_in_ascend_ci__": [
TestFile("ascend/test_mindspore_models.py"),
],
}

suites.update(suite_amd)
Expand Down Expand Up @@ -614,6 +551,30 @@ def _sanity_check_suites(suites):
f"{missing_text}"
)

nonexistent_files = sorted(list(suite_files - disk_files))
nonexistent_text = "\n".join(f'TestFile("{x}"),' for x in nonexistent_files)
assert (
len(nonexistent_files) == 0
), f"Some test files in test suite do not exist on disk:\n{nonexistent_text}"

not_in_ci_files = set(
[test_file.name for test_file in suites.get("__not_in_ci__", [])]
)
in_ci_files = set(
[
test_file.name
for suite_name, suite in suites.items()
if suite_name != "__not_in_ci__"
for test_file in suite
]
)
intersection = not_in_ci_files & in_ci_files
intersection_text = "\n".join(f'TestFile("{x}"),' for x in intersection)
assert len(intersection) == 0, (
f"Some test files are in both `not_in_ci` section and other suites:\n"
f"{intersection_text}"
)


def main():
arg_parser = argparse.ArgumentParser()
Expand Down
Loading