gahow/aituner
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Tighten topology and auto-high validation

Browse Source
This commit is contained in:
Gahow Wang
2026-06-26 20:07:23 +08:00
parent 1dd3eaebaa
commit c8a0f9870e
3 changed files with 128 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

108
tests/test_core_flow.py
View File

@@ -309,6 +309,31 @@ class CoreFlowTests(unittest.TestCase):
self.assertEqual(capped_by_trace.high, 0.7)
self.assertEqual(trace_evidence["effective_ceiling"], 0.7)
low_above_ceiling = study.search.__class__.from_dict(
{
"low": 0.9,
"high": 0.95,
"tolerance": study.search.tolerance,
"max_probes": study.search.max_probes,
"sample_seed": study.search.sample_seed,
"auto_high": {
"enabled": True,
"max_sampling_u": 0.8,
"require_human_confirmation_beyond_trace": True,
},
}
)
unchanged, invalid_evidence = resolve_auto_high_search(
search=low_above_ceiling,
sampling_us=[0.1, 0.9],
)
self.assertEqual(unchanged.low, 0.9)
self.assertEqual(unchanged.high, 0.95)
self.assertEqual(
invalid_evidence["reason"],
"auto_high_ceiling_below_search_low",
)
high_search = study.search.__class__.from_dict(
{
"low": 0.0,
@@ -1531,6 +1556,89 @@ class CoreFlowTests(unittest.TestCase):
"search_high_saturation_requires_parallel_size_evidence",
)
def test_harness_stop_blocks_high_saturation_for_fixed_product_tp_dp_redistribution(
self,
) -> None:
with tempfile.TemporaryDirectory() as tmp:
tmp_path = Path(tmp)
study_path = _write_study_assets(
tmp_path,
engine_overrides={
"base_flags": {
"host": "127.0.0.1",
"port": 8000,
"tensor-parallel-size": 8,
"data-parallel-size": 1,
},
"tunable_flags": ["tensor-parallel-size", "data-parallel-size"],
"topology_constraints": {
"allowed_tensor_parallel_sizes": [1, 2, 4, 8],
"allowed_data_parallel_sizes": [1, 2, 4, 8],
"allowed_tp_dp_products": [8],
"require_tp_dp_product_equals_gpu_count": True,
},
},
)
study = load_study_spec(study_path)
result_path = tmp_path / "trial-0001.json"
result_path.write_text(
json.dumps(
{
"status": "completed",
"best_sampling_u": 0.99609375,
"best_request_rate": 8.0,
"best_pass_rate": 1.0,
"probes": [
{
"threshold": 0.99609375,
"feasible": True,
"payload": {
"request_count": 10,
"pass_rate": 1.0,
"request_rate": 8.0,
"early_stopped": False,
"early_stop_reason": "",
"latency_summary": {"failed_reason_counts": {}},
},
}
],
}
),
encoding="utf-8",
)
state = StudyState(
study_id=study.study_id,
best_trial_id="trial-0001",
best_request_rate=8.0,
best_request_rate_per_gpu=1.0,
trials=[
TrialSummary(
trial_id="trial-0001",
status="completed",
best_request_rate=8.0,
best_request_rate_per_gpu=1.0,
result_path=str(result_path),
config_patch={
"env_patch": {},
"flag_patch": {
"tensor-parallel-size": 8,
"data-parallel-size": 1,
},
},
)
],
)
context = build_harness_context(
study=study,
window_summary={"prompt_tokens_p95": 2048},
state=state,
)
self.assertFalse(context["harness_stop"]["should_stop"])
self.assertEqual(
context["harness_stop"]["reason"],
"search_high_saturation_requires_parallel_size_evidence",
)
def test_harness_guided_first_tp_probe_for_latency_bottleneck(self) -> None:
with tempfile.TemporaryDirectory() as tmp:
tmp_path = Path(tmp)