admin/quant-trader-service
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Improve Trader entry quality training diagnostics

Browse Source
This commit is contained in:
Codex
2026-06-28 00:50:37 +08:00
parent 87849a66a7
commit 340d1dd91b
11 changed files with 1895 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files
training/tests/test_training_contract.py
+207 -3
View File
@@ -14,8 +14,10 @@ if str(TRAINING_ROOT) not in sys.path:
sys.path.insert(0, str(TRAINING_ROOT))
from trader_training.onnx_export import LinearHead, export_heads
from trader_training.entry_feature_screen import _screen_edge_column
from trader_training.io_utils import read_json, write_json
from trader_training.labels import ENTRY_LABEL_METHOD, _path_stats_for_group, build_entry_labels
from trader_training.ofi_feature_experiment import l1_snapshot_diff_ofi_quote
from trader_training.promote import promote_artifact_bundle
from trader_training.replay import build_splits
from trader_training.schemas import FEATURE_ORDER, LATEST_STRESS_SPLIT, MODEL_OUTPUTS, OUTPUT_MAPPING, TRAINING_SPLITS, VALIDATION_LOCKED_SPLIT
@@ -33,6 +35,19 @@ class TrainingContractTest(unittest.TestCase):
self.assertEqual(set(fields), set(OUTPUT_MAPPING[model_name]))
self.assertEqual([f"prediction[{idx}]" for idx in range(len(fields))], [OUTPUT_MAPPING[model_name][field] for field in fields])
def test_entry_feature_screen_prefers_actual_plan_edge(self) -> None:
dataset = pd.DataFrame(
{
"long_expected_net_edge_bps": [20.0],
"short_expected_net_edge_bps": [15.0],
"long_actual_plan_net_edge_bps": [-3.0],
"short_actual_plan_net_edge_bps": [4.0],
}
)
self.assertEqual("long_actual_plan_net_edge_bps", _screen_edge_column(dataset, "LONG"))
self.assertEqual("short_actual_plan_net_edge_bps", _screen_edge_column(dataset, "SHORT"))
def test_split_builder_uses_locked_validation_contract(self) -> None:
with tempfile.TemporaryDirectory() as tmp:
data_root = Path(tmp)
@@ -90,7 +105,7 @@ class TrainingContractTest(unittest.TestCase):
self.assertEqual(120_000, first["time_to_stop_ms"])
self.assertAlmostEqual(-8.0, first["gross_edge_bps"])
def test_entry_label_uses_max_future_edge_not_fixed_target_hit(self) -> None:
def test_entry_label_uses_price_plan_outcome_not_max_future_edge(self) -> None:
with tempfile.TemporaryDirectory() as tmp:
data_root = Path(tmp)
run_root = data_root / "trader-v4" / "runs" / "unit-entry"
@@ -167,11 +182,200 @@ class TrainingContractTest(unittest.TestCase):
labels = pd.read_parquet(run_root / "label" / "entry_labels.parquet")
row = labels[labels["sample_id"].eq("s0") & labels["side"].eq("LONG")].iloc[0]
self.assertEqual(0, row["target_hit"])
self.assertEqual(1, row["entry_target"])
self.assertEqual(0, row["entry_target"])
self.assertEqual(ENTRY_LABEL_METHOD, row["label_method"])
self.assertAlmostEqual(13.5, row["expected_net_edge_bps"], places=6)
self.assertAlmostEqual(-6.5, row["expected_net_edge_bps"], places=6)
self.assertAlmostEqual(row["gross_edge_bps"] - row["cost_bps"], row["expected_net_edge_bps"], places=6)
self.assertAlmostEqual(row["mfe_bps"] - row["cost_bps"], row["max_achievable_net_edge_bps"], places=6)
def test_entry_opportunity_label_keeps_plan_outcome_for_pm(self) -> None:
with tempfile.TemporaryDirectory() as tmp:
data_root = Path(tmp)
run_root = data_root / "trader-v4" / "runs" / "unit-entry-opportunity"
feature_path = run_root / "feature" / "feature_frame.parquet"
replay_path = run_root / "replay" / "replay_1m.parquet"
plan_path = run_root / "label" / "price_plan_context.json"
config_path = data_root / "label_config.json"
feature_path.parent.mkdir(parents=True)
replay_path.parent.mkdir(parents=True)
times = pd.date_range("2026-01-01", periods=5, freq="min", tz="UTC")
pd.DataFrame(
{
"sample_id": ["s0"],
"symbol": "BTC-USDT-PERP",
"event_time": [times[0]],
"open_time_ms": [0],
"split_id": "fit_inner",
"walk_forward_fold": 0,
"data_quality_flag": "OK",
"spread_bps": 1.0,
"spread_rank_24h_pct": 0.1,
"realized_vol_15m_bps": 2.0,
}
).to_parquet(feature_path, index=False)
pd.DataFrame(
{
"event_time": times,
"open_time_ms": np.arange(5, dtype=np.int64) * 60_000,
"symbol": "BTC-USDT-PERP",
"open": [100.0] * 5,
"high": [100.0, 100.05, 100.19, 100.20, 100.0],
"low": [100.0, 99.99, 99.98, 99.97, 100.0],
"close": [100.0] * 5,
"spread_bps": 1.0,
}
).to_parquet(replay_path, index=False)
write_json(
config_path,
{
"entry": {
"max_hold_minutes": 3,
"target_bps": 50.0,
"stop_bps": 50.0,
"min_expected_net_edge_bps": 3.0,
"target_method": "OPPORTUNITY_MFE_V1",
}
},
)
write_json(
plan_path,
{
"pricePlanId": "unit-plan",
"pricePlanConfigHash": "unit-hash",
"targetDistanceBps": 50.0,
"stopDistanceBps": 50.0,
"maxHoldMinutes": 3,
"costBps": 6.5,
"entryLabelMethod": ENTRY_LABEL_METHOD,
"entryTargetMethod": "OPPORTUNITY_MFE_V1",
},
)
build_entry_labels(
Namespace(
data_root=data_root,
run_id="unit-entry-opportunity",
feature_path=feature_path,
replay_path=replay_path,
label_config_path=config_path,
cost_config_path=None,
price_plan_context_path=plan_path,
)
)
labels = pd.read_parquet(run_root / "label" / "entry_labels.parquet")
row = labels[labels["sample_id"].eq("s0") & labels["side"].eq("LONG")].iloc[0]
self.assertEqual(0, row["target_hit"])
self.assertEqual(1, row["entry_target"])
self.assertEqual("OPPORTUNITY_MFE_V1", row["label_method"])
self.assertAlmostEqual(row["mfe_bps"] - row["cost_bps"], row["expected_net_edge_bps"], places=6)
self.assertAlmostEqual(-6.5, row["gross_edge_bps"] - row["cost_bps"], places=6)
def test_entry_quality_label_rejects_untradable_opportunity(self) -> None:
with tempfile.TemporaryDirectory() as tmp:
data_root = Path(tmp)
run_root = data_root / "trader-v4" / "runs" / "unit-entry-quality"
feature_path = run_root / "feature" / "feature_frame.parquet"
replay_path = run_root / "replay" / "replay_1m.parquet"
plan_path = run_root / "label" / "price_plan_context.json"
config_path = data_root / "label_config.json"
feature_path.parent.mkdir(parents=True)
replay_path.parent.mkdir(parents=True)
times = pd.date_range("2026-01-01", periods=5, freq="min", tz="UTC")
pd.DataFrame(
{
"sample_id": ["s0"],
"symbol": "BTC-USDT-PERP",
"event_time": [times[0]],
"open_time_ms": [0],
"split_id": "fit_inner",
"walk_forward_fold": 0,
"data_quality_flag": "OK",
"spread_bps": 1.0,
"spread_rank_24h_pct": 0.1,
"realized_vol_15m_bps": 2.0,
}
).to_parquet(feature_path, index=False)
pd.DataFrame(
{
"event_time": times,
"open_time_ms": np.arange(5, dtype=np.int64) * 60_000,
"symbol": "BTC-USDT-PERP",
"open": [100.0] * 5,
"high": [100.0, 100.05, 100.19, 100.20, 100.0],
"low": [100.0, 99.99, 99.98, 99.97, 100.0],
"close": [100.0] * 5,
"spread_bps": 1.0,
}
).to_parquet(replay_path, index=False)
write_json(
config_path,
{
"entry": {
"max_hold_minutes": 3,
"target_bps": 50.0,
"stop_bps": 50.0,
"min_expected_net_edge_bps": 3.0,
"min_plan_net_edge_bps": 0.0,
"max_entry_mae_bps": 12.0,
"target_method": "OPPORTUNITY_QUALITY_V1",
}
},
)
write_json(
plan_path,
{
"pricePlanId": "unit-plan",
"pricePlanConfigHash": "unit-hash",
"targetDistanceBps": 50.0,
"stopDistanceBps": 50.0,
"maxHoldMinutes": 3,
"costBps": 6.5,
"entryLabelMethod": ENTRY_LABEL_METHOD,
"entryTargetMethod": "OPPORTUNITY_QUALITY_V1",
},
)
build_entry_labels(
Namespace(
data_root=data_root,
run_id="unit-entry-quality",
feature_path=feature_path,
replay_path=replay_path,
label_config_path=config_path,
cost_config_path=None,
price_plan_context_path=plan_path,
)
)
labels = pd.read_parquet(run_root / "label" / "entry_labels.parquet")
row = labels[labels["sample_id"].eq("s0") & labels["side"].eq("LONG")].iloc[0]
self.assertEqual("OPPORTUNITY_QUALITY_V1", row["label_method"])
self.assertGreater(row["expected_net_edge_bps"], 3.0)
self.assertLess(row["actual_plan_net_edge_bps"], 0.0)
self.assertEqual(0, row["entry_target"])
def test_l1_snapshot_diff_ofi_uses_quote_notional_and_signed_ask_side(self) -> None:
bid_part, ask_part = l1_snapshot_diff_ofi_quote(
pd.Series([101.0, 101.0, 100.5]),
pd.Series([2.0, 3.0, 4.0]),
pd.Series([102.0, 101.5, 102.5]),
pd.Series([5.0, 6.0, 7.0]),
pd.Series([100.0, 101.0, 101.0]),
pd.Series([1.5, 2.0, 3.0]),
pd.Series([102.0, 102.0, 101.5]),
pd.Series([4.0, 5.0, 6.0]),
)
self.assertAlmostEqual(202.0, bid_part.iloc[0])
self.assertAlmostEqual(-102.0, ask_part.iloc[0])
self.assertAlmostEqual(101.0, bid_part.iloc[1])
self.assertAlmostEqual(-609.0, ask_part.iloc[1])
self.assertAlmostEqual(-303.0, bid_part.iloc[2])
self.assertAlmostEqual(609.0, ask_part.iloc[2])
def test_exported_onnx_accepts_java_feature_shape(self) -> None:
import onnxruntime as ort