Add hyperparameter and timecost runs

.gitattributes

@@ -58,3 +58,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+*.csv filter=lfs diff=lfs merge=lfs -text
+*.log filter=lfs diff=lfs merge=lfs -text

hyperparameter/c2/arf/arf-c2-20260504_204630/_arf_generate.py

@@ -0,0 +1,93 @@
+import pickle
+import numpy as np
+import pandas as pd
+
+def _bootstrap_from_train(c_csv: str, n_target: int, seed: int = 42) -> pd.DataFrame:
+    """当 arfpy.forge 完全不可用时，从训练 CSV 有放回抽样，保证行数与列对齐。"""
+    src = pd.read_csv(c_csv, encoding="utf-8-sig", low_memory=False)
+    src = src.replace([np.inf, -np.inf], np.nan).dropna(axis=1, how="all")
+    src = src.reset_index(drop=True)
+    if len(src) == 0:
+        raise RuntimeError("ARF fallback: train CSV is empty")
+    return src.sample(n=n_target, replace=True, random_state=seed).reset_index(drop=True)
+
+def _safe_forge(model, n_target: int):
+    # arfpy 在部分分布上会 ZeroDivisionError；n=1 在部分版本会触发
+    # AttributeError（不要用 n=1）。失败返回 None，由外层走 bootstrap。
+    errors = []
+    candidates = []
+    for n_try in (
+        n_target,
+        min(n_target, 8192),
+        min(n_target, 4096),
+        min(n_target, 2048),
+        min(n_target, 1024),
+        min(n_target, 512),
+        256,
+        128,
+        64,
+        32,
+        16,
+        8,
+        2,
+    ):
+        nn = int(n_try)
+        if nn <= 0 or nn in candidates:
+            continue
+        candidates.append(nn)
+    for n_try in candidates:
+        try:
+            out = model.forge(n=n_try).reset_index(drop=True)
+            if len(out) > 0:
+                return out
+        except Exception as e:
+            errors.append(f"n={n_try}: {type(e).__name__}: {e}")
+    print("[ARF] forge failed after retries; last errors:", " | ".join(errors[-4:]))
+    return None
+
+n_target = int(1382)
+c_csv = "/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/train.csv"
+with open("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf_model.pkl", "rb") as f:
+    model = pickle.load(f)
+
+syn = _safe_forge(model, n_target)
+if syn is None or len(syn) == 0:
+    if not c_csv:
+        raise RuntimeError("ARF forge failed and no train csv path for bootstrap fallback")
+    print(f"[ARF] Using train-bootstrap fallback (n={n_target})")
+    syn = _bootstrap_from_train(c_csv, n_target)
+else:
+    if len(syn) > n_target:
+        syn = syn.iloc[:n_target]
+    elif len(syn) < n_target:
+        parts = [syn]
+        tries = 0
+        while sum(len(p) for p in parts) < n_target and tries < 64:
+            tries += 1
+            need = n_target - sum(len(p) for p in parts)
+            chunk = _safe_forge(model, max(need, 2))
+            if chunk is None or len(chunk) == 0:
+                break
+            parts.append(chunk)
+        syn = pd.concat(parts, ignore_index=True).iloc[:n_target]
+    if len(syn) < n_target and c_csv:
+        add_n = n_target - len(syn)
+        add = _bootstrap_from_train(c_csv, add_n, seed=43)
+        syn = pd.concat([syn, add], ignore_index=True).iloc[:n_target]
+
+_ds_id = 'c2'
+if _ds_id == "c19":
+    # 仅 c19：object 列内裸换行会使 pivot 用 csv.reader 统计到的「记录数」大于 DataFrame 行数 → Sw。
+    for _col in syn.columns:
+        if syn[_col].dtype == object:
+            syn[_col] = (
+                syn[_col]
+                .astype(str)
+                .str.replace("\r\n", " ", regex=False)
+                .str.replace("\n", " ", regex=False)
+                .str.replace("\r", " ", regex=False)
+            )
+    syn = syn.iloc[:n_target].reset_index(drop=True)
+
+syn.to_csv("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf-c2-1382-20260504_204635.csv", index=False)
+print(f"[ARF] Generated {len(syn)} rows (requested {n_target}) -> /work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf-c2-1382-20260504_204635.csv")

hyperparameter/c2/arf/arf-c2-20260504_204630/_arf_train.py

@@ -0,0 +1,47 @@
+import os
+import pickle
+import numpy as np
+import pandas as pd
+from arfpy import arf
+
+def _sanitize_for_arf(df: pd.DataFrame) -> pd.DataFrame:
+    """缓解 forge 阶段 scipy.stats.truncnorm / 除零：处理 inf、NaN 与极端尾部。"""
+    df = df.replace([np.inf, -np.inf], np.nan)
+    df = df.dropna(axis=1, how="all")
+    for col in df.select_dtypes(include=[np.number]).columns:
+        med = df[col].median()
+        if pd.isna(med):
+            med = 0.0
+        df[col] = df[col].fillna(med)
+        nu = int(df[col].nunique(dropna=True))
+        if nu <= 1:
+            continue
+        q_low = float(os.environ.get("ARF_CLIP_QUANTILE_LOW", "0.001"))
+        q_high = float(os.environ.get("ARF_CLIP_QUANTILE_HIGH", "0.999"))
+        lo, hi = df[col].quantile(q_low), df[col].quantile(q_high)
+        if pd.notna(lo) and pd.notna(hi) and lo < hi:
+            df[col] = df[col].clip(lo, hi)
+    return df
+
+df = pd.read_csv("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/train.csv")
+df = _sanitize_for_arf(df)
+num_trees = int(os.environ.get("ARF_NUM_TREES", "30"))
+delta = float(os.environ.get("ARF_DELTA", "0"))
+max_iters = int(os.environ.get("ARF_MAX_ITERS", "10"))
+early_stop = (os.environ.get("ARF_EARLY_STOP", "true").strip().lower() in ("1", "true", "yes"))
+verbose = (os.environ.get("ARF_VERBOSE", "true").strip().lower() in ("1", "true", "yes"))
+min_node_size = int(os.environ.get("ARF_MIN_NODE_SIZE", "5"))
+print(f"[ARF] Training on {len(df)} rows, {len(df.columns)} cols")
+print(f"[ARF] Config num_trees={num_trees} delta={delta} max_iters={max_iters} early_stop={early_stop} min_node_size={min_node_size}")
+
+model = arf.arf(x=df, num_trees=num_trees, delta=delta, max_iters=max_iters, early_stop=early_stop, verbose=verbose, min_node_size=min_node_size)
+if hasattr(model, "fit"):
+    model.fit()
+elif hasattr(model, "forde"):
+    model.forde()
+else:
+    raise RuntimeError("arfpy API: no fit() / forde()")
+
+with open("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf_model.pkl", "wb") as f:
+    pickle.dump(model, f)
+print(f"[ARF] Model saved -> /work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf_model.pkl")

hyperparameter/c2/arf/arf-c2-20260504_204630/arf-c2-1382-20260504_204635.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9df075eacde58f79639181162596f36d15c0429e861fd263a248ad87305cee3
+size 41656

hyperparameter/c2/arf/arf-c2-20260504_204630/arf_model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f449587d6d9aa65db19a1f87213e698a6d818c80e0b4054b748ee9688ed33cfa
+size 1423254

hyperparameter/c2/arf/arf-c2-20260504_204630/gen_20260504_204635.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a707a8ca18fca337355b6f940df7056af5af66b06c74b602837a48e55c537ef
+size 2614

hyperparameter/c2/arf/arf-c2-20260504_204630/input_snapshot.json

@@ -0,0 +1,36 @@
+{
+  "dataset_id": "c2",
+  "model": "arf",
+  "inputs": {
+    "train_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-train.csv",
+      "exists": true,
+      "size": 42948,
+      "sha256": "17bc560fa96bd00fb3b526e1e65bc91210b701d0d0a4e8bb9b4c5196cab56def"
+    },
+    "val_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-val.csv",
+      "exists": true,
+      "size": 5349,
+      "sha256": "61e565eca62e65a7dccd9d51039a3170413379e10fc494e25870e7c4294863c9"
+    },
+    "test_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-test.csv",
+      "exists": true,
+      "size": 5448,
+      "sha256": "cbcbb062a1faf5fa44b66c80532baa229e05b94fc42137269761e6c6d84af20a"
+    },
+    "profile_json": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/artifacts/data_core/tabular/c2/c2-dataset_profile.json",
+      "exists": true,
+      "size": 3240,
+      "sha256": "526b7163b2076c93c0bf4638438081ee8a6907065d5b608faa40d1a3dbc2a27b"
+    },
+    "contract_json": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/artifacts/data_core/tabular/c2/c2-dataset_contract_v1.json",
+      "exists": true,
+      "size": 3731,
+      "sha256": "fb595a876054c2ee9b4e10cfe83a5691588de1d25466cbb9d473c18ad3604009"
+    }
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204630/public_gate/normalized_schema_snapshot.json

@@ -0,0 +1,144 @@
+{
+  "dataset_id": "c2",
+  "target_column": "class",
+  "task_type": "classification",
+  "columns": [
+    {
+      "name": "buying",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "med",
+          "high",
+          "low"
+        ]
+      }
+    },
+    {
+      "name": "maint",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "low",
+          "med",
+          "high"
+        ]
+      }
+    },
+    {
+      "name": "doors",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "2",
+          "5more",
+          "3",
+          "4"
+        ]
+      }
+    },
+    {
+      "name": "persons",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "2",
+          "4",
+          "more"
+        ]
+      }
+    },
+    {
+      "name": "lug_boot",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "small",
+          "big",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "safety",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "low",
+          "high",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "class",
+      "role": "target",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "unacc",
+          "good",
+          "acc",
+          "vgood"
+        ]
+      }
+    }
+  ]
+}

hyperparameter/c2/arf/arf-c2-20260504_204630/public_gate/public_gate_report.json

@@ -0,0 +1,37 @@
+{
+  "dataset_id": "c2",
+  "status": "pass",
+  "checks": [
+    {
+      "check_id": "PG001_csv_parse_ok",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG002_split_header_consistent",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG003_profile_header_match",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG004_missing_token_normalized",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG005_semantic_type_validated",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG006_target_defined_and_valid",
+      "status": "pass"
+    }
+  ],
+  "target_column": "class",
+  "task_type": "classification",
+  "input_splits": {
+    "train": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-train.csv",
+    "val": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-val.csv",
+    "test": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-test.csv"
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204630/public_gate/staged_input_manifest.json

@@ -0,0 +1,149 @@
+{
+  "dataset_id": "c2",
+  "target_column": "class",
+  "task_type": "classification",
+  "train_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/train.csv",
+  "val_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/val.csv",
+  "test_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/test.csv",
+  "features_json": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/staged_features.json",
+  "public_gate_report": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/public_gate/public_gate_report.json",
+  "column_schema": [
+    {
+      "name": "buying",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "med",
+          "high",
+          "low"
+        ]
+      }
+    },
+    {
+      "name": "maint",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "low",
+          "med",
+          "high"
+        ]
+      }
+    },
+    {
+      "name": "doors",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "2",
+          "5more",
+          "3",
+          "4"
+        ]
+      }
+    },
+    {
+      "name": "persons",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "2",
+          "4",
+          "more"
+        ]
+      }
+    },
+    {
+      "name": "lug_boot",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "small",
+          "big",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "safety",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "low",
+          "high",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "class",
+      "role": "target",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "unacc",
+          "good",
+          "acc",
+          "vgood"
+        ]
+      }
+    }
+  ]
+}

hyperparameter/c2/arf/arf-c2-20260504_204630/run_config.json

@@ -0,0 +1,43 @@
+{
+  "schema_version": 1,
+  "recorded_at": "2026-05-04T20:46:30",
+  "dataset_id": "c2",
+  "model": "arf",
+  "work_dir": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630",
+  "dataset_source_requested": "new",
+  "dataset_source_resolved": "new",
+  "cli_args": {
+    "model": "arf",
+    "dataset": "c2",
+    "dataset_source": "new",
+    "train": true,
+    "generate": true,
+    "num_rows": 0,
+    "epochs": null,
+    "output_dir": null,
+    "model_dir": null,
+    "work_dir": null,
+    "resume": false,
+    "no_stats": false
+  },
+  "resolved": {
+    "num_rows": 1382,
+    "model_path": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf_model.pkl",
+    "output_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf-c2-1382-20260504_204635.csv"
+  },
+  "input_artifacts": {
+    "public_gate_report": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/public_gate/public_gate_report.json",
+    "public_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/public_gate/staged_input_manifest.json",
+    "model_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/arf/model_input_manifest.json",
+    "train_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/train.csv",
+    "features_json": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/staged_features.json",
+    "target_column": "class",
+    "task_type": "classification"
+  },
+  "env_overrides": {
+    "ARF_DELTA": "0.01",
+    "ARF_MAX_ITERS": "3",
+    "ARF_MIN_NODE_SIZE": "7",
+    "ARF_NUM_TREES": "11"
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204630/runtime_result.json

@@ -0,0 +1,27 @@
+{
+  "dataset_id": "c2",
+  "model": "arf",
+  "run_id": "arf-c2-20260504_204630",
+  "public_gate_status": "pass",
+  "adapter_ready_status": "pass",
+  "train_status": "success",
+  "generate_status": "success",
+  "reason_code": null,
+  "reason_detail": null,
+  "artifacts": {
+    "synthetic_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf-c2-1382-20260504_204635.csv",
+    "model_path": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/arf_model.pkl"
+  },
+  "timings": {
+    "train": {
+      "started_at": "2026-05-04T20:46:30",
+      "ended_at": "2026-05-04T20:46:35",
+      "duration_sec": 4.953
+    },
+    "generate": {
+      "started_at": "2026-05-04T20:46:35",
+      "ended_at": "2026-05-04T20:46:37",
+      "duration_sec": 2.615
+    }
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204630/staged/arf/adapter_report.json

@@ -0,0 +1,7 @@
+{
+  "adapter_ready_status": "pass",
+  "adapter_fail_reason_code": null,
+  "adapter_fail_detail": null,
+  "adapter_transforms_applied": [],
+  "model_input_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/arf/model_input_manifest.json"
+}

hyperparameter/c2/arf/arf-c2-20260504_204630/staged/arf/adapter_transforms_applied.json

@@ -0,0 +1 @@
+[]

hyperparameter/c2/arf/arf-c2-20260504_204630/staged/arf/model_input_manifest.json

@@ -0,0 +1,151 @@
+{
+  "dataset_id": "c2",
+  "model": "arf",
+  "target_column": "class",
+  "task_type": "classification",
+  "column_schema": [
+    {
+      "name": "buying",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "med",
+          "high",
+          "low"
+        ]
+      }
+    },
+    {
+      "name": "maint",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "low",
+          "med",
+          "high"
+        ]
+      }
+    },
+    {
+      "name": "doors",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "2",
+          "5more",
+          "3",
+          "4"
+        ]
+      }
+    },
+    {
+      "name": "persons",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "2",
+          "4",
+          "more"
+        ]
+      }
+    },
+    {
+      "name": "lug_boot",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "small",
+          "big",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "safety",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "low",
+          "high",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "class",
+      "role": "target",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "unacc",
+          "good",
+          "acc",
+          "vgood"
+        ]
+      }
+    }
+  ],
+  "public_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/public_gate/staged_input_manifest.json",
+  "train_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/train.csv",
+  "val_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/val.csv",
+  "test_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/test.csv",
+  "features_json": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/staged/public/staged_features.json",
+  "public_gate_report": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204630/public_gate/public_gate_report.json"
+}

hyperparameter/c2/arf/arf-c2-20260504_204630/staged/public/staged_features.json

@@ -0,0 +1,37 @@
+[
+  {
+    "feature_name": "buying",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "maint",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "doors",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "persons",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "lug_boot",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "safety",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "class",
+    "data_type": "categorical",
+    "is_target": true
+  }
+]

hyperparameter/c2/arf/arf-c2-20260504_204630/staged/public/test.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b48114a7d0bc5bd9a07920f903c8d4aba8bf98bf2a66a050da03588b0245ca73
+size 5273

hyperparameter/c2/arf/arf-c2-20260504_204630/staged/public/train.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4aed00c2c2b3f88a55a7ebff31b2e1b5e0e32fb0a7267e0b9d2779cd23e434dd
+size 41565

hyperparameter/c2/arf/arf-c2-20260504_204630/staged/public/val.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26e90c1170a57a14c05832ac88027722b1f3848f9662c7c09ef7c93dcba4cc01
+size 5176

hyperparameter/c2/arf/arf-c2-20260504_204630/train_20260504_204630.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:253c9d5a82654e53a261a511acde522fe48d5affeed38350eb4ef2f3fefb61ca
+size 1002

hyperparameter/c2/arf/arf-c2-20260504_204810/_arf_generate.py

@@ -0,0 +1,93 @@
+import pickle
+import numpy as np
+import pandas as pd
+
+def _bootstrap_from_train(c_csv: str, n_target: int, seed: int = 42) -> pd.DataFrame:
+    """当 arfpy.forge 完全不可用时，从训练 CSV 有放回抽样，保证行数与列对齐。"""
+    src = pd.read_csv(c_csv, encoding="utf-8-sig", low_memory=False)
+    src = src.replace([np.inf, -np.inf], np.nan).dropna(axis=1, how="all")
+    src = src.reset_index(drop=True)
+    if len(src) == 0:
+        raise RuntimeError("ARF fallback: train CSV is empty")
+    return src.sample(n=n_target, replace=True, random_state=seed).reset_index(drop=True)
+
+def _safe_forge(model, n_target: int):
+    # arfpy 在部分分布上会 ZeroDivisionError；n=1 在部分版本会触发
+    # AttributeError（不要用 n=1）。失败返回 None，由外层走 bootstrap。
+    errors = []
+    candidates = []
+    for n_try in (
+        n_target,
+        min(n_target, 8192),
+        min(n_target, 4096),
+        min(n_target, 2048),
+        min(n_target, 1024),
+        min(n_target, 512),
+        256,
+        128,
+        64,
+        32,
+        16,
+        8,
+        2,
+    ):
+        nn = int(n_try)
+        if nn <= 0 or nn in candidates:
+            continue
+        candidates.append(nn)
+    for n_try in candidates:
+        try:
+            out = model.forge(n=n_try).reset_index(drop=True)
+            if len(out) > 0:
+                return out
+        except Exception as e:
+            errors.append(f"n={n_try}: {type(e).__name__}: {e}")
+    print("[ARF] forge failed after retries; last errors:", " | ".join(errors[-4:]))
+    return None
+
+n_target = int(1382)
+c_csv = "/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/train.csv"
+with open("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf_model.pkl", "rb") as f:
+    model = pickle.load(f)
+
+syn = _safe_forge(model, n_target)
+if syn is None or len(syn) == 0:
+    if not c_csv:
+        raise RuntimeError("ARF forge failed and no train csv path for bootstrap fallback")
+    print(f"[ARF] Using train-bootstrap fallback (n={n_target})")
+    syn = _bootstrap_from_train(c_csv, n_target)
+else:
+    if len(syn) > n_target:
+        syn = syn.iloc[:n_target]
+    elif len(syn) < n_target:
+        parts = [syn]
+        tries = 0
+        while sum(len(p) for p in parts) < n_target and tries < 64:
+            tries += 1
+            need = n_target - sum(len(p) for p in parts)
+            chunk = _safe_forge(model, max(need, 2))
+            if chunk is None or len(chunk) == 0:
+                break
+            parts.append(chunk)
+        syn = pd.concat(parts, ignore_index=True).iloc[:n_target]
+    if len(syn) < n_target and c_csv:
+        add_n = n_target - len(syn)
+        add = _bootstrap_from_train(c_csv, add_n, seed=43)
+        syn = pd.concat([syn, add], ignore_index=True).iloc[:n_target]
+
+_ds_id = 'c2'
+if _ds_id == "c19":
+    # 仅 c19：object 列内裸换行会使 pivot 用 csv.reader 统计到的「记录数」大于 DataFrame 行数 → Sw。
+    for _col in syn.columns:
+        if syn[_col].dtype == object:
+            syn[_col] = (
+                syn[_col]
+                .astype(str)
+                .str.replace("\r\n", " ", regex=False)
+                .str.replace("\n", " ", regex=False)
+                .str.replace("\r", " ", regex=False)
+            )
+    syn = syn.iloc[:n_target].reset_index(drop=True)
+
+syn.to_csv("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf-c2-1382-20260504_204814.csv", index=False)
+print(f"[ARF] Generated {len(syn)} rows (requested {n_target}) -> /work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf-c2-1382-20260504_204814.csv")

hyperparameter/c2/arf/arf-c2-20260504_204810/_arf_train.py

@@ -0,0 +1,47 @@
+import os
+import pickle
+import numpy as np
+import pandas as pd
+from arfpy import arf
+
+def _sanitize_for_arf(df: pd.DataFrame) -> pd.DataFrame:
+    """缓解 forge 阶段 scipy.stats.truncnorm / 除零：处理 inf、NaN 与极端尾部。"""
+    df = df.replace([np.inf, -np.inf], np.nan)
+    df = df.dropna(axis=1, how="all")
+    for col in df.select_dtypes(include=[np.number]).columns:
+        med = df[col].median()
+        if pd.isna(med):
+            med = 0.0
+        df[col] = df[col].fillna(med)
+        nu = int(df[col].nunique(dropna=True))
+        if nu <= 1:
+            continue
+        q_low = float(os.environ.get("ARF_CLIP_QUANTILE_LOW", "0.001"))
+        q_high = float(os.environ.get("ARF_CLIP_QUANTILE_HIGH", "0.999"))
+        lo, hi = df[col].quantile(q_low), df[col].quantile(q_high)
+        if pd.notna(lo) and pd.notna(hi) and lo < hi:
+            df[col] = df[col].clip(lo, hi)
+    return df
+
+df = pd.read_csv("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/train.csv")
+df = _sanitize_for_arf(df)
+num_trees = int(os.environ.get("ARF_NUM_TREES", "30"))
+delta = float(os.environ.get("ARF_DELTA", "0"))
+max_iters = int(os.environ.get("ARF_MAX_ITERS", "10"))
+early_stop = (os.environ.get("ARF_EARLY_STOP", "true").strip().lower() in ("1", "true", "yes"))
+verbose = (os.environ.get("ARF_VERBOSE", "true").strip().lower() in ("1", "true", "yes"))
+min_node_size = int(os.environ.get("ARF_MIN_NODE_SIZE", "5"))
+print(f"[ARF] Training on {len(df)} rows, {len(df.columns)} cols")
+print(f"[ARF] Config num_trees={num_trees} delta={delta} max_iters={max_iters} early_stop={early_stop} min_node_size={min_node_size}")
+
+model = arf.arf(x=df, num_trees=num_trees, delta=delta, max_iters=max_iters, early_stop=early_stop, verbose=verbose, min_node_size=min_node_size)
+if hasattr(model, "fit"):
+    model.fit()
+elif hasattr(model, "forde"):
+    model.forde()
+else:
+    raise RuntimeError("arfpy API: no fit() / forde()")
+
+with open("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf_model.pkl", "wb") as f:
+    pickle.dump(model, f)
+print(f"[ARF] Model saved -> /work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf_model.pkl")

hyperparameter/c2/arf/arf-c2-20260504_204810/arf-c2-1382-20260504_204814.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32362e2f23866523d5f892582e5cc2b208df5d33725c0707ec379ea9b04a5c30
+size 41763

hyperparameter/c2/arf/arf-c2-20260504_204810/arf_model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1659d887626798c3c856fa91f789ac96630081bb395eb91b42fc1ab00ca73c50
+size 1286384

hyperparameter/c2/arf/arf-c2-20260504_204810/gen_20260504_204814.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5f59afe1f1e0abc5eb11ad94446e3d6db029e2104e0466ac23dbfc3606ac649
+size 2615

hyperparameter/c2/arf/arf-c2-20260504_204810/input_snapshot.json

@@ -0,0 +1,36 @@
+{
+  "dataset_id": "c2",
+  "model": "arf",
+  "inputs": {
+    "train_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-train.csv",
+      "exists": true,
+      "size": 42948,
+      "sha256": "17bc560fa96bd00fb3b526e1e65bc91210b701d0d0a4e8bb9b4c5196cab56def"
+    },
+    "val_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-val.csv",
+      "exists": true,
+      "size": 5349,
+      "sha256": "61e565eca62e65a7dccd9d51039a3170413379e10fc494e25870e7c4294863c9"
+    },
+    "test_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-test.csv",
+      "exists": true,
+      "size": 5448,
+      "sha256": "cbcbb062a1faf5fa44b66c80532baa229e05b94fc42137269761e6c6d84af20a"
+    },
+    "profile_json": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/artifacts/data_core/tabular/c2/c2-dataset_profile.json",
+      "exists": true,
+      "size": 3240,
+      "sha256": "526b7163b2076c93c0bf4638438081ee8a6907065d5b608faa40d1a3dbc2a27b"
+    },
+    "contract_json": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/artifacts/data_core/tabular/c2/c2-dataset_contract_v1.json",
+      "exists": true,
+      "size": 3731,
+      "sha256": "fb595a876054c2ee9b4e10cfe83a5691588de1d25466cbb9d473c18ad3604009"
+    }
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204810/public_gate/normalized_schema_snapshot.json

@@ -0,0 +1,144 @@
+{
+  "dataset_id": "c2",
+  "target_column": "class",
+  "task_type": "classification",
+  "columns": [
+    {
+      "name": "buying",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "med",
+          "high",
+          "low"
+        ]
+      }
+    },
+    {
+      "name": "maint",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "low",
+          "med",
+          "high"
+        ]
+      }
+    },
+    {
+      "name": "doors",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "2",
+          "5more",
+          "3",
+          "4"
+        ]
+      }
+    },
+    {
+      "name": "persons",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "2",
+          "4",
+          "more"
+        ]
+      }
+    },
+    {
+      "name": "lug_boot",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "small",
+          "big",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "safety",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "low",
+          "high",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "class",
+      "role": "target",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "unacc",
+          "good",
+          "acc",
+          "vgood"
+        ]
+      }
+    }
+  ]
+}

hyperparameter/c2/arf/arf-c2-20260504_204810/public_gate/public_gate_report.json

@@ -0,0 +1,37 @@
+{
+  "dataset_id": "c2",
+  "status": "pass",
+  "checks": [
+    {
+      "check_id": "PG001_csv_parse_ok",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG002_split_header_consistent",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG003_profile_header_match",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG004_missing_token_normalized",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG005_semantic_type_validated",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG006_target_defined_and_valid",
+      "status": "pass"
+    }
+  ],
+  "target_column": "class",
+  "task_type": "classification",
+  "input_splits": {
+    "train": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-train.csv",
+    "val": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-val.csv",
+    "test": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-test.csv"
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204810/public_gate/staged_input_manifest.json

@@ -0,0 +1,149 @@
+{
+  "dataset_id": "c2",
+  "target_column": "class",
+  "task_type": "classification",
+  "train_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/train.csv",
+  "val_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/val.csv",
+  "test_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/test.csv",
+  "features_json": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/staged_features.json",
+  "public_gate_report": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/public_gate/public_gate_report.json",
+  "column_schema": [
+    {
+      "name": "buying",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "med",
+          "high",
+          "low"
+        ]
+      }
+    },
+    {
+      "name": "maint",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "low",
+          "med",
+          "high"
+        ]
+      }
+    },
+    {
+      "name": "doors",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "2",
+          "5more",
+          "3",
+          "4"
+        ]
+      }
+    },
+    {
+      "name": "persons",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "2",
+          "4",
+          "more"
+        ]
+      }
+    },
+    {
+      "name": "lug_boot",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "small",
+          "big",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "safety",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "low",
+          "high",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "class",
+      "role": "target",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "unacc",
+          "good",
+          "acc",
+          "vgood"
+        ]
+      }
+    }
+  ]
+}

hyperparameter/c2/arf/arf-c2-20260504_204810/run_config.json

@@ -0,0 +1,46 @@
+{
+  "schema_version": 1,
+  "recorded_at": "2026-05-04T20:48:10",
+  "dataset_id": "c2",
+  "model": "arf",
+  "work_dir": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810",
+  "dataset_source_requested": "new",
+  "dataset_source_resolved": "new",
+  "cli_args": {
+    "model": "arf",
+    "dataset": "c2",
+    "dataset_source": "new",
+    "train": true,
+    "generate": true,
+    "num_rows": 0,
+    "epochs": null,
+    "output_dir": null,
+    "model_dir": null,
+    "work_dir": null,
+    "resume": false,
+    "no_stats": false
+  },
+  "resolved": {
+    "num_rows": 1382,
+    "model_path": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf_model.pkl",
+    "output_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf-c2-1382-20260504_204814.csv"
+  },
+  "input_artifacts": {
+    "public_gate_report": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/public_gate/public_gate_report.json",
+    "public_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/public_gate/staged_input_manifest.json",
+    "model_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/arf/model_input_manifest.json",
+    "train_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/train.csv",
+    "features_json": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/staged_features.json",
+    "target_column": "class",
+    "task_type": "classification"
+  },
+  "env_overrides": {
+    "ARF_CLIP_QUANTILE_HIGH": "0.999",
+    "ARF_CLIP_QUANTILE_LOW": "0.001",
+    "ARF_DELTA": "0",
+    "ARF_EARLY_STOP": "true",
+    "ARF_MAX_ITERS": "5",
+    "ARF_MIN_NODE_SIZE": "7",
+    "ARF_NUM_TREES": "10"
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204810/runtime_result.json

@@ -0,0 +1,27 @@
+{
+  "dataset_id": "c2",
+  "model": "arf",
+  "run_id": "arf-c2-20260504_204810",
+  "public_gate_status": "pass",
+  "adapter_ready_status": "pass",
+  "train_status": "success",
+  "generate_status": "success",
+  "reason_code": null,
+  "reason_detail": null,
+  "artifacts": {
+    "synthetic_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf-c2-1382-20260504_204814.csv",
+    "model_path": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/arf_model.pkl"
+  },
+  "timings": {
+    "train": {
+      "started_at": "2026-05-04T20:48:10",
+      "ended_at": "2026-05-04T20:48:14",
+      "duration_sec": 4.906
+    },
+    "generate": {
+      "started_at": "2026-05-04T20:48:14",
+      "ended_at": "2026-05-04T20:48:17",
+      "duration_sec": 2.707
+    }
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204810/staged/arf/adapter_report.json

@@ -0,0 +1,7 @@
+{
+  "adapter_ready_status": "pass",
+  "adapter_fail_reason_code": null,
+  "adapter_fail_detail": null,
+  "adapter_transforms_applied": [],
+  "model_input_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/arf/model_input_manifest.json"
+}

hyperparameter/c2/arf/arf-c2-20260504_204810/staged/arf/adapter_transforms_applied.json

@@ -0,0 +1 @@
+[]

hyperparameter/c2/arf/arf-c2-20260504_204810/staged/arf/model_input_manifest.json

@@ -0,0 +1,151 @@
+{
+  "dataset_id": "c2",
+  "model": "arf",
+  "target_column": "class",
+  "task_type": "classification",
+  "column_schema": [
+    {
+      "name": "buying",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "med",
+          "high",
+          "low"
+        ]
+      }
+    },
+    {
+      "name": "maint",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "low",
+          "med",
+          "high"
+        ]
+      }
+    },
+    {
+      "name": "doors",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "2",
+          "5more",
+          "3",
+          "4"
+        ]
+      }
+    },
+    {
+      "name": "persons",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "2",
+          "4",
+          "more"
+        ]
+      }
+    },
+    {
+      "name": "lug_boot",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "small",
+          "big",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "safety",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "low",
+          "high",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "class",
+      "role": "target",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "unacc",
+          "good",
+          "acc",
+          "vgood"
+        ]
+      }
+    }
+  ],
+  "public_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/public_gate/staged_input_manifest.json",
+  "train_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/train.csv",
+  "val_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/val.csv",
+  "test_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/test.csv",
+  "features_json": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/staged/public/staged_features.json",
+  "public_gate_report": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204810/public_gate/public_gate_report.json"
+}

hyperparameter/c2/arf/arf-c2-20260504_204810/staged/public/staged_features.json

@@ -0,0 +1,37 @@
+[
+  {
+    "feature_name": "buying",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "maint",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "doors",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "persons",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "lug_boot",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "safety",
+    "data_type": "categorical",
+    "is_target": false
+  },
+  {
+    "feature_name": "class",
+    "data_type": "categorical",
+    "is_target": true
+  }
+]

hyperparameter/c2/arf/arf-c2-20260504_204810/staged/public/test.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b48114a7d0bc5bd9a07920f903c8d4aba8bf98bf2a66a050da03588b0245ca73
+size 5273

hyperparameter/c2/arf/arf-c2-20260504_204810/staged/public/train.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4aed00c2c2b3f88a55a7ebff31b2e1b5e0e32fb0a7267e0b9d2779cd23e434dd
+size 41565

hyperparameter/c2/arf/arf-c2-20260504_204810/staged/public/val.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26e90c1170a57a14c05832ac88027722b1f3848f9662c7c09ef7c93dcba4cc01
+size 5176

hyperparameter/c2/arf/arf-c2-20260504_204810/train_20260504_204810.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:766114cb37dc6a845fe6f50743760528206e5f07dc8cbb7a67c5069424376baa
+size 999

hyperparameter/c2/arf/arf-c2-20260504_204831/_arf_generate.py

@@ -0,0 +1,93 @@
+import pickle
+import numpy as np
+import pandas as pd
+
+def _bootstrap_from_train(c_csv: str, n_target: int, seed: int = 42) -> pd.DataFrame:
+    """当 arfpy.forge 完全不可用时，从训练 CSV 有放回抽样，保证行数与列对齐。"""
+    src = pd.read_csv(c_csv, encoding="utf-8-sig", low_memory=False)
+    src = src.replace([np.inf, -np.inf], np.nan).dropna(axis=1, how="all")
+    src = src.reset_index(drop=True)
+    if len(src) == 0:
+        raise RuntimeError("ARF fallback: train CSV is empty")
+    return src.sample(n=n_target, replace=True, random_state=seed).reset_index(drop=True)
+
+def _safe_forge(model, n_target: int):
+    # arfpy 在部分分布上会 ZeroDivisionError；n=1 在部分版本会触发
+    # AttributeError（不要用 n=1）。失败返回 None，由外层走 bootstrap。
+    errors = []
+    candidates = []
+    for n_try in (
+        n_target,
+        min(n_target, 8192),
+        min(n_target, 4096),
+        min(n_target, 2048),
+        min(n_target, 1024),
+        min(n_target, 512),
+        256,
+        128,
+        64,
+        32,
+        16,
+        8,
+        2,
+    ):
+        nn = int(n_try)
+        if nn <= 0 or nn in candidates:
+            continue
+        candidates.append(nn)
+    for n_try in candidates:
+        try:
+            out = model.forge(n=n_try).reset_index(drop=True)
+            if len(out) > 0:
+                return out
+        except Exception as e:
+            errors.append(f"n={n_try}: {type(e).__name__}: {e}")
+    print("[ARF] forge failed after retries; last errors:", " | ".join(errors[-4:]))
+    return None
+
+n_target = int(1382)
+c_csv = "/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/public/train.csv"
+with open("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf_model.pkl", "rb") as f:
+    model = pickle.load(f)
+
+syn = _safe_forge(model, n_target)
+if syn is None or len(syn) == 0:
+    if not c_csv:
+        raise RuntimeError("ARF forge failed and no train csv path for bootstrap fallback")
+    print(f"[ARF] Using train-bootstrap fallback (n={n_target})")
+    syn = _bootstrap_from_train(c_csv, n_target)
+else:
+    if len(syn) > n_target:
+        syn = syn.iloc[:n_target]
+    elif len(syn) < n_target:
+        parts = [syn]
+        tries = 0
+        while sum(len(p) for p in parts) < n_target and tries < 64:
+            tries += 1
+            need = n_target - sum(len(p) for p in parts)
+            chunk = _safe_forge(model, max(need, 2))
+            if chunk is None or len(chunk) == 0:
+                break
+            parts.append(chunk)
+        syn = pd.concat(parts, ignore_index=True).iloc[:n_target]
+    if len(syn) < n_target and c_csv:
+        add_n = n_target - len(syn)
+        add = _bootstrap_from_train(c_csv, add_n, seed=43)
+        syn = pd.concat([syn, add], ignore_index=True).iloc[:n_target]
+
+_ds_id = 'c2'
+if _ds_id == "c19":
+    # 仅 c19：object 列内裸换行会使 pivot 用 csv.reader 统计到的「记录数」大于 DataFrame 行数 → Sw。
+    for _col in syn.columns:
+        if syn[_col].dtype == object:
+            syn[_col] = (
+                syn[_col]
+                .astype(str)
+                .str.replace("\r\n", " ", regex=False)
+                .str.replace("\n", " ", regex=False)
+                .str.replace("\r", " ", regex=False)
+            )
+    syn = syn.iloc[:n_target].reset_index(drop=True)
+
+syn.to_csv("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf-c2-1382-20260504_204837.csv", index=False)
+print(f"[ARF] Generated {len(syn)} rows (requested {n_target}) -> /work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf-c2-1382-20260504_204837.csv")

hyperparameter/c2/arf/arf-c2-20260504_204831/_arf_train.py

@@ -0,0 +1,47 @@
+import os
+import pickle
+import numpy as np
+import pandas as pd
+from arfpy import arf
+
+def _sanitize_for_arf(df: pd.DataFrame) -> pd.DataFrame:
+    """缓解 forge 阶段 scipy.stats.truncnorm / 除零：处理 inf、NaN 与极端尾部。"""
+    df = df.replace([np.inf, -np.inf], np.nan)
+    df = df.dropna(axis=1, how="all")
+    for col in df.select_dtypes(include=[np.number]).columns:
+        med = df[col].median()
+        if pd.isna(med):
+            med = 0.0
+        df[col] = df[col].fillna(med)
+        nu = int(df[col].nunique(dropna=True))
+        if nu <= 1:
+            continue
+        q_low = float(os.environ.get("ARF_CLIP_QUANTILE_LOW", "0.001"))
+        q_high = float(os.environ.get("ARF_CLIP_QUANTILE_HIGH", "0.999"))
+        lo, hi = df[col].quantile(q_low), df[col].quantile(q_high)
+        if pd.notna(lo) and pd.notna(hi) and lo < hi:
+            df[col] = df[col].clip(lo, hi)
+    return df
+
+df = pd.read_csv("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/public/train.csv")
+df = _sanitize_for_arf(df)
+num_trees = int(os.environ.get("ARF_NUM_TREES", "30"))
+delta = float(os.environ.get("ARF_DELTA", "0"))
+max_iters = int(os.environ.get("ARF_MAX_ITERS", "10"))
+early_stop = (os.environ.get("ARF_EARLY_STOP", "true").strip().lower() in ("1", "true", "yes"))
+verbose = (os.environ.get("ARF_VERBOSE", "true").strip().lower() in ("1", "true", "yes"))
+min_node_size = int(os.environ.get("ARF_MIN_NODE_SIZE", "5"))
+print(f"[ARF] Training on {len(df)} rows, {len(df.columns)} cols")
+print(f"[ARF] Config num_trees={num_trees} delta={delta} max_iters={max_iters} early_stop={early_stop} min_node_size={min_node_size}")
+
+model = arf.arf(x=df, num_trees=num_trees, delta=delta, max_iters=max_iters, early_stop=early_stop, verbose=verbose, min_node_size=min_node_size)
+if hasattr(model, "fit"):
+    model.fit()
+elif hasattr(model, "forde"):
+    model.forde()
+else:
+    raise RuntimeError("arfpy API: no fit() / forde()")
+
+with open("/work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf_model.pkl", "wb") as f:
+    pickle.dump(model, f)
+print(f"[ARF] Model saved -> /work/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf_model.pkl")

hyperparameter/c2/arf/arf-c2-20260504_204831/arf-c2-1382-20260504_204837.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:220d45777265ecddbcf2c8d2a46cc14c2a8062fd66ccb2e1f47165c8825787e4
+size 41587

hyperparameter/c2/arf/arf-c2-20260504_204831/arf_model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31491a21e917fb29cead65a033642886ad50b01a3a901ea0d1e811c2874bce8d
+size 5074463

hyperparameter/c2/arf/arf-c2-20260504_204831/gen_20260504_204837.log

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abf47bcac1aa8a34b442950b08f11ab7c8cd88b0d6d41c69448c1b3002858b29
+size 2615

hyperparameter/c2/arf/arf-c2-20260504_204831/input_snapshot.json

@@ -0,0 +1,36 @@
+{
+  "dataset_id": "c2",
+  "model": "arf",
+  "inputs": {
+    "train_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-train.csv",
+      "exists": true,
+      "size": 42948,
+      "sha256": "17bc560fa96bd00fb3b526e1e65bc91210b701d0d0a4e8bb9b4c5196cab56def"
+    },
+    "val_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-val.csv",
+      "exists": true,
+      "size": 5349,
+      "sha256": "61e565eca62e65a7dccd9d51039a3170413379e10fc494e25870e7c4294863c9"
+    },
+    "test_csv": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-test.csv",
+      "exists": true,
+      "size": 5448,
+      "sha256": "cbcbb062a1faf5fa44b66c80532baa229e05b94fc42137269761e6c6d84af20a"
+    },
+    "profile_json": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/artifacts/data_core/tabular/c2/c2-dataset_profile.json",
+      "exists": true,
+      "size": 3240,
+      "sha256": "526b7163b2076c93c0bf4638438081ee8a6907065d5b608faa40d1a3dbc2a27b"
+    },
+    "contract_json": {
+      "path": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/artifacts/data_core/tabular/c2/c2-dataset_contract_v1.json",
+      "exists": true,
+      "size": 3731,
+      "sha256": "fb595a876054c2ee9b4e10cfe83a5691588de1d25466cbb9d473c18ad3604009"
+    }
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204831/public_gate/normalized_schema_snapshot.json

@@ -0,0 +1,144 @@
+{
+  "dataset_id": "c2",
+  "target_column": "class",
+  "task_type": "classification",
+  "columns": [
+    {
+      "name": "buying",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "med",
+          "high",
+          "low"
+        ]
+      }
+    },
+    {
+      "name": "maint",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "low",
+          "med",
+          "high"
+        ]
+      }
+    },
+    {
+      "name": "doors",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "2",
+          "5more",
+          "3",
+          "4"
+        ]
+      }
+    },
+    {
+      "name": "persons",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "2",
+          "4",
+          "more"
+        ]
+      }
+    },
+    {
+      "name": "lug_boot",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "small",
+          "big",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "safety",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "low",
+          "high",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "class",
+      "role": "target",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "unacc",
+          "good",
+          "acc",
+          "vgood"
+        ]
+      }
+    }
+  ]
+}

hyperparameter/c2/arf/arf-c2-20260504_204831/public_gate/public_gate_report.json

@@ -0,0 +1,37 @@
+{
+  "dataset_id": "c2",
+  "status": "pass",
+  "checks": [
+    {
+      "check_id": "PG001_csv_parse_ok",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG002_split_header_consistent",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG003_profile_header_match",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG004_missing_token_normalized",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG005_semantic_type_validated",
+      "status": "pass"
+    },
+    {
+      "check_id": "PG006_target_defined_and_valid",
+      "status": "pass"
+    }
+  ],
+  "target_column": "class",
+  "task_type": "classification",
+  "input_splits": {
+    "train": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-train.csv",
+    "val": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-val.csv",
+    "test": "/data/jialinzhang/SynthesizePipeline-server/DatasetNew/c2/c2-test.csv"
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204831/public_gate/staged_input_manifest.json

@@ -0,0 +1,149 @@
+{
+  "dataset_id": "c2",
+  "target_column": "class",
+  "task_type": "classification",
+  "train_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/public/train.csv",
+  "val_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/public/val.csv",
+  "test_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/public/test.csv",
+  "features_json": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/public/staged_features.json",
+  "public_gate_report": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/public_gate/public_gate_report.json",
+  "column_schema": [
+    {
+      "name": "buying",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "med",
+          "high",
+          "low"
+        ]
+      }
+    },
+    {
+      "name": "maint",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "vhigh",
+          "low",
+          "med",
+          "high"
+        ]
+      }
+    },
+    {
+      "name": "doors",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "2",
+          "5more",
+          "3",
+          "4"
+        ]
+      }
+    },
+    {
+      "name": "persons",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "2",
+          "4",
+          "more"
+        ]
+      }
+    },
+    {
+      "name": "lug_boot",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "small",
+          "big",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "safety",
+      "role": "feature",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 3,
+        "unique_ratio": 0.002171,
+        "example_values": [
+          "low",
+          "high",
+          "med"
+        ]
+      }
+    },
+    {
+      "name": "class",
+      "role": "target",
+      "semantic_type": "categorical",
+      "nullable": false,
+      "missing_tokens": [],
+      "parse_format": null,
+      "impute_strategy": "mode",
+      "profile_stats": {
+        "missing_rate": 0.0,
+        "unique_count": 4,
+        "unique_ratio": 0.002894,
+        "example_values": [
+          "unacc",
+          "good",
+          "acc",
+          "vgood"
+        ]
+      }
+    }
+  ]
+}

hyperparameter/c2/arf/arf-c2-20260504_204831/run_config.json

@@ -0,0 +1,46 @@
+{
+  "schema_version": 1,
+  "recorded_at": "2026-05-04T20:48:31",
+  "dataset_id": "c2",
+  "model": "arf",
+  "work_dir": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831",
+  "dataset_source_requested": "new",
+  "dataset_source_resolved": "new",
+  "cli_args": {
+    "model": "arf",
+    "dataset": "c2",
+    "dataset_source": "new",
+    "train": true,
+    "generate": true,
+    "num_rows": 0,
+    "epochs": null,
+    "output_dir": null,
+    "model_dir": null,
+    "work_dir": null,
+    "resume": false,
+    "no_stats": false
+  },
+  "resolved": {
+    "num_rows": 1382,
+    "model_path": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf_model.pkl",
+    "output_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf-c2-1382-20260504_204837.csv"
+  },
+  "input_artifacts": {
+    "public_gate_report": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/public_gate/public_gate_report.json",
+    "public_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/public_gate/staged_input_manifest.json",
+    "model_manifest": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/arf/model_input_manifest.json",
+    "train_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/public/train.csv",
+    "features_json": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/staged/public/staged_features.json",
+    "target_column": "class",
+    "task_type": "classification"
+  },
+  "env_overrides": {
+    "ARF_CLIP_QUANTILE_HIGH": "0.999",
+    "ARF_CLIP_QUANTILE_LOW": "0.001",
+    "ARF_DELTA": "0",
+    "ARF_EARLY_STOP": "true",
+    "ARF_MAX_ITERS": "10",
+    "ARF_MIN_NODE_SIZE": "5",
+    "ARF_NUM_TREES": "30"
+  }
+}

hyperparameter/c2/arf/arf-c2-20260504_204831/runtime_result.json

@@ -0,0 +1,27 @@
+{
+  "dataset_id": "c2",
+  "model": "arf",
+  "run_id": "arf-c2-20260504_204831",
+  "public_gate_status": "pass",
+  "adapter_ready_status": "pass",
+  "train_status": "success",
+  "generate_status": "success",
+  "reason_code": null,
+  "reason_detail": null,
+  "artifacts": {
+    "synthetic_csv": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf-c2-1382-20260504_204837.csv",
+    "model_path": "/data/jialinzhang/SynthesizePipeline-server/output-Benchmark-trainonly-v1/c2/arf/arf-c2-20260504_204831/arf_model.pkl"
+  },
+  "timings": {
+    "train": {
+      "started_at": "2026-05-04T20:48:31",
+      "ended_at": "2026-05-04T20:48:37",
+      "duration_sec": 6.228
+    },
+    "generate": {
+      "started_at": "2026-05-04T20:48:37",
+      "ended_at": "2026-05-04T20:48:40",
+      "duration_sec": 2.878
+    }
+  }
+}