feat(fge): simplify the training workflow (#62)

* initial commit

* update example

Author: xuyxu

CHANGELOG.rst

@@ -4,6 +4,7 @@ Changelog
 Ver 0.1.*
 ---------
 
+* |Enhancement| |API| Simplify the training workflow of :class:`FastGeometricClassifier` and :class:`FastGeometricRegressor` | `@xuyxu <https://github.com/xuyxu>`__
 * |Feature| |API| Support TensorBoard logging in :meth:`set_logger` | `@zzzzwj <https://github.com/zzzzwj>`__
 * |Enhancement| |API| Add ``use_reduction_sum`` parameter for :meth:`fit` of Gradient Boosting | `@xuyxu <https://github.com/xuyxu>`__
 * |Feature| |API| Improve the functionality of :meth:`evaluate` and :meth:`predict` | `@xuyxu <https://github.com/xuyxu>`__

docs/parameters.rst

@@ -147,21 +147,6 @@ Reference:
     T. Garipov, P. Izmailov, D. Podoprikhin et al., Loss Surfaces, Mode
     Connectivity, and Fast Ensembling of DNNs, NeurIPS, 2018.
 
-Notice that unlike all ensembles above, using fast geometric ensemble (FGE) is
-**a two-staged process**. Concretely, you first need to call :meth:`fit` to
-build a dummy base estimator that will be used to generate ensembles. Second,
-you need to call :meth:`ensemble` to generate real base estimators in the
-ensemble. The pipeline is shown in the following code snippet:
-
-.. code:: python
-
-    model = FastGeometricClassifier(**ensemble_related_args)
-    estimator = model.fit(train_loader, **base_estimator_related_args)  # train the base estimator
-    model.ensemble(estimator, train_loader, **fge_related_args)  # generate the ensemble using the base estimator
-
-You can refer to scripts in `examples <https://github.com/xuyxu/Ensemble-Pytorch/tree/master/examples>`__ for
-a detailed example.
-
 FastGeometricClassifier
 ***********************
 

examples/fast_geometric_ensemble_cifar10_resnet18.py

@@ -161,16 +161,9 @@ def forward(self, x):
         train_loader,
         epochs=epochs,
         test_loader=test_loader,
-    )
-
-    # Ensemble
-    model.ensemble(
-        estimator,
-        train_loader,
         cycle=4,
         lr_1=5e-2,
         lr_2=5e-4,
-        test_loader=test_loader,
     )
 
     # Evaluate