Updated everything to pytorch 0.41. And Gitignore.

.gitignore

@@ -0,0 +1,4 @@
+*~
+*checkpoint.ipynb
+__pycache__
+*.pyc

EEG-BCI-Project/cross_validation.py

@@ -63,9 +63,9 @@ def train_model(model, criterion, optimizer, scheduler,train_input, train_target
     """
     # Augment the data if necessary and make Variables
     train_input, train_target = augment_train(train_input, train_target, n_augmentation, 0.1, 2 , 0.2,verbose)
-    train_input, train_target = Variable(train_input), Variable(train_target)
-    if(save_loss):
-        val_input,val_target = Variable(val_input),Variable(val_target)
+    #train_input, train_target = Variable(train_input), Variable(train_target) # No longer necessary in Pytorch 0.4
+    #if(save_loss): # No longer necessary in Pytorch 0.4
+    #    val_input,val_target = Variable(val_input),Variable(val_target) # No longer necessary in Pytorch 0.4
 
     for e in range(n_epochs):
         scheduler.step() # decrease the learning rate
@@ -88,10 +88,10 @@ def train_model(model, criterion, optimizer, scheduler,train_input, train_target
                 optimizer.step()
         # save loss data, if save_loss is true
         if(save_loss):
-            tr_loss.append(criterion(model(train_input),train_target).data[0])
-            val_loss.append(criterion(model(val_input),val_target).data[0])
-            tr_err.append(evaluate_error(model,train_input.data,train_target.data))
-            val_err.append(evaluate_error(model,val_input.data,val_target.data))
+            tr_loss.append(criterion(model(train_input),train_target).item())#item and detach from pytorch 0.4
+            val_loss.append(criterion(model(val_input),val_target).item())#item and detach from pytorch 0.4
+            tr_err.append(evaluate_error(model,train_input.detach(),train_target.detach()).item()) #item and detach from pytorch 0.4
+            val_err.append(evaluate_error(model,val_input.detach(),val_target.detach()).item()) #item and detach from pytorch 0.4
     if(verbose == 1):
         print("Training ended successfully after {} epochs".format(n_epochs))
 
@@ -103,13 +103,14 @@ def evaluate_error(model, data_input, data_target):
             data_target -> the correct labels for the input data
     Outputs: the number of missclassified samples
     """
-    data_input, data_target = Variable(data_input,volatile=True),Variable(data_target,volatile=True)
-    nb_errors = 0
+    #data_input, data_target = Variable(data_input,volatile=True),Variable(data_target,volatile=True)# No longer necessary in Pytorch 0.4
+    with torch.no_grad(): 
+        nb_errors = 0
 
-    output = model(data_input)
-    nb_errors += (output.max(1)[1] != data_target).long().data.sum() # count the number of samples in the output with labels different than in the target
+        output = model(data_input)
+        nb_errors += (output.max(1)[1] != data_target).long().detach().sum() # count the number of samples in the output with labels different than in the target #detach is for pytorch 0.4
 
-    return nb_errors/data_input.size(0) # take the mean, to get a fraction of missclassified samples
+        return nb_errors.float()/data_input.size(0) # take the mean, to get a fraction of missclassified samples #pytorch 0.4 need to be float to be able to divide.
 
 def cross_validate(model,criterion,optimizer,scheduler,dataset,target,k_fold,n_epochs=250,batch_size = 50,n_augmentation =0,verbose=0):
     """

EEG-BCI-Project/data_bci/labels_data_set_iv.txt

@@ -0,0 +1,100 @@
+1
+0
+0
+0
+1
+0
+0
+0
+1
+1
+1
+0
+0
+1
+1
+0
+0
+0
+0
+0
+0
+0
+0
+1
+0
+1
+1
+1
+0
+1
+0
+0
+1
+1
+0
+0
+1
+0
+1
+1
+1
+1
+0
+0
+0
+1
+0
+0
+1
+1
+1
+1
+1
+0
+1
+1
+1
+1
+0
+1
+1
+1
+0
+1
+0
+0
+1
+0
+0
+1
+0
+1
+1
+0
+0
+0
+0
+0
+1
+1
+0
+1
+0
+1
+1
+1
+0
+1
+0
+1
+1
+0
+1
+0
+1
+1
+0
+1
+1
+0

EEG-BCI-Project/final_500t.txt

@@ -0,0 +1 @@
+[["Linear baseline", " train ", 0.0917721539735794], ["Linear baseline", " test ", 0.3499999940395355], ["Convolutional baseline", " train ", 0.2405063360929489], ["Convolutional baseline", " test ", 0.27000001072883606], ["EEG Net (2018)", " train ", 0.06962025165557861], ["EEG Net (2018)", " test ", 0.1899999976158142], ["ShallowConvNet", " train ", 0.09810126572847366], ["ShallowConvNet", " test ", 0.4099999964237213]]

EEG-BCI-Project/plot_CV_results.py

@@ -60,6 +60,7 @@
     plt.xlabel('Epoch')
     plt.ylabel('Loss (Cross entropy)')
     for i in range(0,len(lines)-2*n_last,2):
+        print(lines[i])
         model = lines[i][0]
         tr_loss = lines[i][2][2:]
         val_loss = lines[i+1][2][2:]

EEG-BCI-Project/test.py

@@ -22,6 +22,10 @@
 
 """Load the data and standardize"""
 import dlc_bci as bci
+import torch
+
+dtype = torch.float
+device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 print("Loading the data and standardizing...")
 train_input ,train_target = bci.load( root = './data_bci',one_khz=one_khz)
 test_input,test_target = bci.load(root='./data_bci', train = False, one_khz=one_khz)
@@ -59,6 +63,7 @@
 
 import cross_validation as cv #Also used for the train_model, even without cross validation_data
 import json
+
 dump_final = []
 if(cross_val):
     dump = []
@@ -84,8 +89,8 @@
     final_te_error = cv.evaluate_error(model,test_input,test_target)
     print("Train error = {} ; Test error = {} ".format(final_tr_error,final_te_error))
 
-    dump_final.append((model.name(), " train ", final_tr_error))
-    dump_final.append((model.name(), " test " , final_te_error))
+    dump_final.append((model.name(), " train ", final_tr_error.item()))
+    dump_final.append((model.name(), " test " , final_te_error.item()))
 
 
 file = open('final_'+str(n_time_points)+'t.txt','w+')