makes the optimizer optional

README.md

@@ -235,12 +235,6 @@ model = NeuralNet(data_set, cnn_reg, model_type='3d', task='reg',
                   metrics_exporters=[OutputExporter(out)],
                   cuda=False)
 
-# change the optimizer (optional)
-model.optimizer = optim.SGD(model.net.parameters(),
-                            lr=0.001,
-                            momentum=0.9,
-                            weight_decay=0.005)
-
 # do the prediction
 model.test()
 ```

deeprank/learn/NeuralNet.py

@@ -223,16 +223,9 @@ def __init__(self, data_set, model,
         elif self.cuda:
             self.net = self.net.cuda()
 
-        # set the optimizer
-        #self.optimizer = optim.SGD(self.net.parameters(),
-        #                           lr=0.005,
-        #                           momentum=0.9,
-        #                           weight_decay=0.001)
-        self.optimizer = optim.AdamW(self.net.parameters(),
-                                     lr=0.005,
-                                     weight_decay=0.001)
-        if self.pretrained_model:
-            self.load_optimizer_params()
+        # set the optimizer to None in the beginning.
+        # if the user is going to train a model, then he must set it.
+        self.optimizer = None
 
         # ------------------------------------------
         # print
@@ -298,6 +291,13 @@ def train(self,
             save_model (str, optional): 'best' or 'all', save only the
                 best model or all models.
         """
+
+        if self.optimizer is None:
+            if self.pretrained_model is not None:
+                self.load_optimizer_params()
+            else:
+                raise RuntimeError("no optimizer set, cannot train")
+
         logger.info(f'\n: Batch Size: {train_batch_size}')
         if self.cuda:
             logger.info(f': NGPU      : {self.ngpu}')
@@ -407,6 +407,18 @@ def load_model_params(self):
 
     def load_optimizer_params(self):
         """Get optimizer parameters from a saved model."""
+
+        # guess the optimizer
+        if 'momentum' in self.state['optimizer']['param_groups']:
+            self.optimizer = optim.SGD(self.net.parameters(),
+                                       lr=0.005,
+                                       momentum=0.9,
+                                       weight_decay=0.001)
+        else:
+            self.optimizer = optim.AdamW(self.net.parameters(),
+                                         lr=0.005,
+                                         weight_decay=0.001)
+
         self.optimizer.load_state_dict(self.state['optimizer'])
 
     def load_nn_params(self):

test/test_learn.py

@@ -22,9 +22,12 @@
 
 
 
-def test_predict_without_target():
-    feature_modules = ["test.feature.feature1", "test.feature.feature2"]
+def test_predict():
     target_modules = []
+    feature_modules = [
+        'deeprank.features.atomic_contacts',
+        'deeprank.features.neighbour_profile',
+        'deeprank.features.accessibility']
 
     atomic_densities = {'C': 1.7, 'N': 1.55, 'O': 1.52, 'S': 1.8}
     grid_info = { 
@@ -37,14 +40,8 @@ def test_predict_without_target():
 
     variants = [PdbVariantSelection("101m", "A", 10, valine, cysteine,
                                     protein_accession="P02144", protein_residue_number=10,
-                                    ),
-                PdbVariantSelection("101m", "A", 8, glutamine, cysteine,
-                                    protein_accession="P02144",
-                                    ),
-                PdbVariantSelection("101m", "A", 9, glutamine, cysteine,
-                                    protein_accession="P02144", protein_residue_number=9,
                                     )]
-    augmentation = 5
+    augmentation = 2
 
     work_dir_path = mkdtemp()
     try:
@@ -72,18 +69,9 @@ def test_predict_without_target():
         metrics_directory = os.path.join(work_dir_path, "runs")
 
         neural_net = NeuralNet(dataset, cnn_class, model_type='3d',task='class',
+                               pretrained_model="test/data/models/best_valid_model.pth.tar",
                                cuda=False, metrics_exporters=[OutputExporter(metrics_directory),
                                                               TensorboardBinaryClassificationExporter(metrics_directory)])
-
-        neural_net.optimizer = optim.SGD(neural_net.net.parameters(),
-                                         lr=0.001,
-                                         momentum=0.9,
-                                         weight_decay=0.005)
-
-        neural_net.state = {}
-        neural_net.state["task"] = "class"
-        neural_net.state["criterion"] = neural_net.criterion
-
         neural_net.test()
     finally:
         rmtree(work_dir_path)