From 629cab2399889c8af8976c26144ecaf48490ecf0 Mon Sep 17 00:00:00 2001
From: Yufeng He <40085740+he-yufeng@users.noreply.github.com>
Date: Sat, 23 May 2026 19:37:38 +0800
Subject: [PATCH] fix: allow disabling LightGBM early stopping

---
 qlib/contrib/model/gbdt.py                  | 12 +++-
 qlib/contrib/model/highfreq_gdbt_model.py   |  7 ++-
 tests/model/test_lightgbm_early_stopping.py | 65 +++++++++++++++++++++
 3 files changed, 79 insertions(+), 5 deletions(-)
 create mode 100644 tests/model/test_lightgbm_early_stopping.py

diff --git a/qlib/contrib/model/gbdt.py b/qlib/contrib/model/gbdt.py
index 22c29cd4997..b3c10998dec 100644
--- a/qlib/contrib/model/gbdt.py
+++ b/qlib/contrib/model/gbdt.py
@@ -68,19 +68,25 @@ def fit(
             evals_result = {}  # in case of unsafety of Python default values
         ds_l = self._prepare_data(dataset, reweighter)
         ds, names = list(zip(*ds_l))
-        early_stopping_callback = lgb.early_stopping(
-            self.early_stopping_rounds if early_stopping_rounds is None else early_stopping_rounds
+        early_stopping_rounds = (
+            self.early_stopping_rounds
+            if early_stopping_rounds is None
+            else early_stopping_rounds
         )
+        callbacks = []
+        if early_stopping_rounds is not None:
+            callbacks.append(lgb.early_stopping(early_stopping_rounds))
         # NOTE: if you encounter error here. Please upgrade your lightgbm
         verbose_eval_callback = lgb.log_evaluation(period=verbose_eval)
         evals_result_callback = lgb.record_evaluation(evals_result)
+        callbacks.extend([verbose_eval_callback, evals_result_callback])
         self.model = lgb.train(
             self.params,
             ds[0],  # training dataset
             num_boost_round=self.num_boost_round if num_boost_round is None else num_boost_round,
             valid_sets=ds,
             valid_names=names,
-            callbacks=[early_stopping_callback, verbose_eval_callback, evals_result_callback],
+            callbacks=callbacks,
             **kwargs,
         )
         for k in names:
diff --git a/qlib/contrib/model/highfreq_gdbt_model.py b/qlib/contrib/model/highfreq_gdbt_model.py
index ad0641136f2..e2f09846d8c 100644
--- a/qlib/contrib/model/highfreq_gdbt_model.py
+++ b/qlib/contrib/model/highfreq_gdbt_model.py
@@ -124,16 +124,19 @@ def fit(
         if evals_result is None:
             evals_result = dict()
         dtrain, dvalid = self._prepare_data(dataset)
-        early_stopping_callback = lgb.early_stopping(early_stopping_rounds)
+        callbacks = []
+        if early_stopping_rounds is not None:
+            callbacks.append(lgb.early_stopping(early_stopping_rounds))
         verbose_eval_callback = lgb.log_evaluation(period=verbose_eval)
         evals_result_callback = lgb.record_evaluation(evals_result)
+        callbacks.extend([verbose_eval_callback, evals_result_callback])
         self.model = lgb.train(
             self.params,
             dtrain,
             num_boost_round=num_boost_round,
             valid_sets=[dtrain, dvalid],
             valid_names=["train", "valid"],
-            callbacks=[early_stopping_callback, verbose_eval_callback, evals_result_callback],
+            callbacks=callbacks,
         )
         evals_result["train"] = list(evals_result["train"].values())[0]
         evals_result["valid"] = list(evals_result["valid"].values())[0]
diff --git a/tests/model/test_lightgbm_early_stopping.py b/tests/model/test_lightgbm_early_stopping.py
new file mode 100644
index 00000000000..90606331a3f
--- /dev/null
+++ b/tests/model/test_lightgbm_early_stopping.py
@@ -0,0 +1,65 @@
+from qlib.contrib.model import gbdt, highfreq_gdbt_model
+
+
+def test_lgb_model_skips_none_early_stopping(monkeypatch):
+    early_stopping_calls = []
+    train_kwargs = {}
+    model = gbdt.LGBModel(early_stopping_rounds=None)
+
+    monkeypatch.setattr(
+        model,
+        "_prepare_data",
+        lambda dataset, reweighter=None: [("train-dataset", "train")],
+    )
+    monkeypatch.setattr(
+        gbdt.lgb,
+        "early_stopping",
+        lambda rounds: early_stopping_calls.append(rounds) or "early",
+    )
+    monkeypatch.setattr(gbdt.lgb, "log_evaluation", lambda period: "log")
+    monkeypatch.setattr(gbdt.lgb, "record_evaluation", lambda evals_result: "record")
+
+    def train(*args, **kwargs):
+        train_kwargs.update(kwargs)
+        return object()
+
+    monkeypatch.setattr(gbdt.lgb, "train", train)
+
+    model.fit(object(), evals_result={"train": {}})
+
+    assert early_stopping_calls == []
+    assert train_kwargs["callbacks"] == ["log", "record"]
+
+
+def test_hflgb_model_skips_none_early_stopping(monkeypatch):
+    early_stopping_calls = []
+    train_kwargs = {}
+    model = highfreq_gdbt_model.HFLGBModel()
+
+    monkeypatch.setattr(
+        model, "_prepare_data", lambda dataset: ("train-dataset", "valid-dataset")
+    )
+    monkeypatch.setattr(
+        highfreq_gdbt_model.lgb,
+        "early_stopping",
+        lambda rounds: early_stopping_calls.append(rounds) or "early",
+    )
+    monkeypatch.setattr(highfreq_gdbt_model.lgb, "log_evaluation", lambda period: "log")
+    monkeypatch.setattr(
+        highfreq_gdbt_model.lgb, "record_evaluation", lambda evals_result: "record"
+    )
+
+    def train(*args, **kwargs):
+        train_kwargs.update(kwargs)
+        return object()
+
+    monkeypatch.setattr(highfreq_gdbt_model.lgb, "train", train)
+
+    model.fit(
+        object(),
+        early_stopping_rounds=None,
+        evals_result={"train": {"loss": [1.0]}, "valid": {"loss": [1.1]}},
+    )
+
+    assert early_stopping_calls == []
+    assert train_kwargs["callbacks"] == ["log", "record"]