remove an argument

src/train.py

@@ -152,7 +152,6 @@
                     0.5,
                     shuffle=False,
                     k_iwae=num_samples,
-                    model_name=args.net,
                 )
         if itr % 100 == 0 and args.save:
             torch.save({

src/utils.py

@@ -64,7 +64,6 @@ def evaluate_hetvae(
     sample_tp=0.5,
     shuffle=False,
     k_iwae=1,
-    model_name=None,
     device='cuda',
 ):
     torch.manual_seed(seed=0)
@@ -157,12 +156,6 @@ def get_mimiciii_data(batch_size, test_batch_size=5, filter_anomalies=True):
                 data_mean.append(0)
                 data_std.append(1)
 
-        # filtering
-        observed_vals[observed_vals > hth] = 0.0
-        observed_vals[observed_vals < lth] = 0.0
-        observed_mask[observed_vals > hth] = 0
-        observed_mask[observed_vals < lth] = 0
-
         # normalizing
         observed_vals = (observed_vals - data_mean) / data_std
         observed_vals[observed_mask == 0] = 0
@@ -189,8 +182,8 @@ def get_mimiciii_data(batch_size, test_batch_size=5, filter_anomalies=True):
         total_dataset, train_size=0.8, random_state=42, shuffle=True
     )
     # for interpolation, we dont need a non-overlapping validation set as
-    # we can condition on different set of time points from same dataset to
-    # create a distinct example
+    # we can condition on different set of time points from same set to
+    # create distinct examples
     _, val_data = model_selection.train_test_split(
         train_data, train_size=0.8, random_state=11, shuffle=True
     )