Saving the model

docs/source/whats_new.rst

@@ -25,6 +25,8 @@ Enhancements
 - Adding second deployment of the documentation (:gh:`374` by `Bruno Aristimunha`_)
 - Adding Parallel evaluation for :func:`moabb.evaluations.WithinSessionEvaluation` , :func:`moabb.evaluations.CrossSessionEvaluation` (:gh:`364` by `Bruno Aristimunha`_)
 - Add example with VirtualReality BrainInvaders dataset (:gh:`393` by `Gregoire Cattan`_ and `Pedro L. C. Rodrigues`_)
+- Adding saving option for the models (:gh:`401` by `Bruno Aristimunha`_ and `Igor Carrara`_)
+- Adding example to load different type of models (:gh:`401` by `Bruno Aristimunha`_ and `Igor Carrara`_)
 - Add resting state paradigm with dataset and example (:gh:`400` by `Gregoire Cattan`_ and `Pedro L. C. Rodrigues`_)
 
 Bugs
@@ -33,6 +35,7 @@ Bugs
 - Restore 3 subject from Cho2017 (:gh:`392` by `Igor Carrara`_ and `Sylvain Chevallier`_)
 - Correct downloading with VirtualReality BrainInvaders dataset (:gh:`393` by `Gregoire Cattan`_)
 - Rename event `substraction` to `subtraction` in :func:`moabb.dataset.Shin2017B` (:gh:`397` by `Pierre Guetschel`_)
+- Fixing issue with parallel evaluation (:gh:`401` by `Bruno Aristimunha`_ and `Igor Carrara`_)
 
 API changes
 ~~~~~~~~~~~

examples/advanced_examples/plot_grid_search_withinsession.py

@@ -145,7 +145,7 @@
         path,
         "GridSearch_WithinSession",
         "001-2014",
-        "subject1",
+        "1",
         "session_E",
         "GridSearchEN",
         "Grid_Search_WithinSession.pkl",
@@ -165,7 +165,7 @@
         path,
         "GridSearch_WithinSession",
         "001-2014",
-        "subject1",
+        "1",
         "session_T",
         "GridSearchEN",
         "Grid_Search_WithinSession.pkl",

examples/plot_load_model.py

@@ -0,0 +1,123 @@
+"""
+==============================================
+Load Model (Scikit, Pytorch, Keras) with MOABB
+==============================================
+
+This example shows how to use load the pretrained pipeline in MOABB.
+"""
+# Authors: Igor Carrara <[email protected]>
+#
+# License: BSD (3-clause)
+
+from pickle import load
+
+import keras
+import torch
+from braindecode import EEGClassifier
+from braindecode.models import EEGInception
+from scikeras.wrappers import KerasClassifier
+from sklearn.pipeline import Pipeline
+from skorch.callbacks import EarlyStopping, EpochScoring
+from skorch.dataset import ValidSplit
+
+from moabb import set_log_level
+from moabb.pipelines.features import StandardScaler_Epoch
+from moabb.pipelines.utils_pytorch import BraindecodeDatasetLoader, InputShapeSetterEEG
+from moabb.utils import setup_seed
+
+
+set_log_level("info")
+
+###############################################################################
+# In this example, we will use the results computed by the following examples
+#
+# - plot_benchmark_
+# - plot_benchmark_braindecode_
+# - plot_benchmark_DL_
+# ---------------------
+
+# Set up reproducibility of Tensorflow and PyTorch
+setup_seed(42)
+
+###############################################################################
+# Loading the Scikit-learn pipelines
+
+with open(
+    "./results/Models_WithinSession/Zhou 2016/1/session_0/CSP + SVM/fitted_model_best.pkl",
+    "rb",
+) as pickle_file:
+    CSP_SVM_Trained = load(pickle_file)
+
+###############################################################################
+# Loading the Keras model
+# We load the single Keras model, if we want we can set in the exact same pipeline.
+
+model_Keras = keras.models.load_model(
+    "./results/Models_WithinSession/001-2014/1/session_E/Keras_DeepConvNet/kerasdeepconvnet_fitted_model_best.h5"
+)
+# Now we need to instantiate a new SciKeras object since we only saved the Keras model
+Keras_DeepConvNet_Trained = KerasClassifier(model_Keras)
+# Create the pipelines
+
+
+pipes_keras = Pipeline(
+    [
+        ("StandardScaler_Epoch", StandardScaler_Epoch),
+        ("Keras_DeepConvNet_Trained", Keras_DeepConvNet_Trained),
+    ]
+)
+
+
+###############################################################################
+# Loading the PyTorch model
+
+# Set EEG Inception model
+model = EEGInception(in_channels=22, n_classes=2)
+
+# Hyperparameter
+LEARNING_RATE = 0.0001
+WEIGHT_DECAY = 0
+BATCH_SIZE = 64
+SEED = 42
+VERBOSE = 1
+EPOCH = 2
+PATIENCE = 3
+
+# Define a Skorch classifier
+clf = EEGClassifier(
+    module=model,
+    criterion=torch.nn.CrossEntropyLoss,
+    optimizer=torch.optim.Adam,
+    optimizer__lr=LEARNING_RATE,
+    batch_size=BATCH_SIZE,
+    max_epochs=EPOCH,
+    train_split=ValidSplit(0.2, random_state=SEED),
+    callbacks=[
+        EarlyStopping(monitor="valid_loss", patience=PATIENCE),
+        EpochScoring(
+            scoring="accuracy", on_train=True, name="train_acc", lower_is_better=False
+        ),
+        EpochScoring(
+            scoring="accuracy", on_train=False, name="valid_acc", lower_is_better=False
+        ),
+        InputShapeSetterEEG(
+            params_list=["in_channels", "input_window_samples", "n_classes"],
+        ),
+    ],
+    verbose=VERBOSE,  # Not printing the results for each epoch
+)
+
+clf.initialize()
+
+f_params = "./results/Models_CrossSession/001-2014/1/braindecode_EEGInception/EEGInception_fitted_best_model.pkl"
+f_optimizer = "./results/Models_CrossSession/001-2014/1/braindecode_EEGInception/EEGInception_fitted_best_optim.pkl"
+f_history = "./results/Models_CrossSession/001-2014/1/braindecode_EEGInception/EEGInception_fitted_best_history.json"
+
+clf.load_params(f_params=f_params, f_optimizer=f_optimizer, f_history=f_history)
+
+
+# Create the dataset
+create_dataset = BraindecodeDatasetLoader(drop_last_window=False)
+
+# Create the pipelines
+pipes_pytorch = Pipeline([("Braindecode_dataset", create_dataset), ("EEGInception", clf)])

moabb/benchmark.py

@@ -40,6 +40,7 @@ def benchmark(  # noqa: C901
     overwrite=False,
     output="./benchmark/",
     n_jobs=-1,
+    n_jobs_evaluation=1,
     plot=False,
     contexts=None,
     include_datasets=None,
@@ -85,6 +86,9 @@ def benchmark(  # noqa: C901
         Folder to store the analysis results
     n_jobs: int
         Number of threads to use for running parallel jobs
+    n_jobs_evaluation: int, default=1
+        Number of jobs for evaluation, processing in parallel the within session,
+        cross-session or cross-subject.
     plot: bool
         Plot results after computing
     contexts: str
@@ -172,7 +176,8 @@ def benchmark(  # noqa: C901
                     datasets=d,
                     random_state=42,
                     hdf5_path=results,
-                    n_jobs=1,
+                    n_jobs=n_jobs,
+                    n_jobs_evaluation=n_jobs_evaluation,
                     overwrite=overwrite,
                     return_epochs=True,
                 )
@@ -192,6 +197,7 @@ def benchmark(  # noqa: C901
                     random_state=42,
                     hdf5_path=results,
                     n_jobs=n_jobs,
+                    n_jobs_evaluation=n_jobs_evaluation,
                     overwrite=overwrite,
                 )
                 paradigm_results = context.process(

moabb/evaluations/__init__.py

@@ -10,3 +10,4 @@
     CrossSubjectEvaluation,
     WithinSessionEvaluation,
 )
+from .utils import create_save_path, save_model_cv, save_model_list