Remove stratified group k fold (#101)

open-spaced-repetition · Apr 2, 2024 · e81789e · e81789e
1 parent 11db4ec
commit e81789e
Show file tree

Hide file tree

Showing 2 changed files with 31 additions and 53 deletions.
diff --git a/pyproject.toml b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "FSRS-Optimizer"
-version = "4.27.2"
+version = "4.27.3"
 readme = "README.md"
 dependencies = [
     "matplotlib>=3.7.0",

diff --git a/src/fsrs_optimizer/fsrs_optimizer.py b/src/fsrs_optimizer/fsrs_optimizer.py
@@ -16,7 +16,7 @@
 from torch import Tensor
 from torch.utils.data import Dataset
 from torch.nn.utils.rnn import pad_sequence
-from sklearn.model_selection import StratifiedGroupKFold, TimeSeriesSplit
+from sklearn.model_selection import TimeSeriesSplit
 from sklearn.metrics import root_mean_squared_error, mean_absolute_error, r2_score
 from scipy.optimize import minimize
 from itertools import accumulate
@@ -977,7 +977,6 @@ def train(
         self,
         lr: float = 4e-2,
         n_epoch: int = 5,
-        n_splits: int = 5,
         batch_size: int = 512,
         verbose: bool = True,
         split_by_time: bool = False,
@@ -993,59 +992,38 @@ def train(
 
         w = []
         plots = []
-        if n_splits > 1:
-            if split_by_time:
-                tscv = TimeSeriesSplit(n_splits=n_splits)
-                self.dataset.sort_values(by=["review_time"], inplace=True)
-                for i, (train_index, test_index) in enumerate(tscv.split(self.dataset)):
-                    if verbose:
-                        tqdm.write(f"TRAIN: {len(train_index)} TEST: {len(test_index)}")
-                    train_set = self.dataset.iloc[train_index].copy()
-                    test_set = self.dataset.iloc[test_index].copy()
-                    trainer = Trainer(
-                        train_set,
-                        test_set,
-                        self.init_w,
-                        n_epoch=n_epoch,
-                        lr=lr,
-                        batch_size=batch_size,
-                    )
-                    w.append(trainer.train(verbose=verbose))
-                    self.w = w[-1]
-                    self.evaluate()
-                    metrics, figures = self.calibration_graph(
-                        self.dataset.iloc[test_index]
-                    )
-                    for j, f in enumerate(figures):
-                        f.savefig(f"graph_{j}_test_{i}.png")
-                        plt.close(f)
-                    if verbose:
-                        print(metrics)
-                        plots.append(trainer.plot())
-            else:
-                sgkf = StratifiedGroupKFold(n_splits=n_splits)
-                for train_index, test_index in sgkf.split(
-                    self.dataset, self.dataset["i"], self.dataset["group"]
-                ):
-                    if verbose:
-                        tqdm.write(f"TRAIN: {len(train_index)} TEST: {len(test_index)}")
-                    train_set = self.dataset.iloc[train_index].copy()
-                    test_set = self.dataset.iloc[test_index].copy()
-                    trainer = Trainer(
-                        train_set,
-                        test_set,
-                        self.init_w,
-                        n_epoch=n_epoch,
-                        lr=lr,
-                        batch_size=batch_size,
-                    )
-                    w.append(trainer.train(verbose=verbose))
-                    if verbose:
-                        plots.append(trainer.plot())
+        if split_by_time:
+            tscv = TimeSeriesSplit(n_splits=5)
+            self.dataset.sort_values(by=["review_time"], inplace=True)
+            for i, (train_index, test_index) in enumerate(tscv.split(self.dataset)):
+                if verbose:
+                    tqdm.write(f"TRAIN: {len(train_index)} TEST: {len(test_index)}")
+                train_set = self.dataset.iloc[train_index].copy()
+                test_set = self.dataset.iloc[test_index].copy()
+                trainer = Trainer(
+                    train_set,
+                    test_set,
+                    self.init_w,
+                    n_epoch=n_epoch,
+                    lr=lr,
+                    batch_size=batch_size,
+                )
+                w.append(trainer.train(verbose=verbose))
+                self.w = w[-1]
+                self.evaluate()
+                metrics, figures = self.calibration_graph(
+                    self.dataset.iloc[test_index]
+                )
+                for j, f in enumerate(figures):
+                    f.savefig(f"graph_{j}_test_{i}.png")
+                    plt.close(f)
+                if verbose:
+                    print(metrics)
+                    plots.append(trainer.plot())
         else:
             trainer = Trainer(
                 self.dataset,
-                self.dataset,
+                None,
                 self.init_w,
                 n_epoch=n_epoch,
                 lr=lr,