- small changes

few_shot_tests.py

@@ -55,6 +55,9 @@ def parser():
     parser.add_argument('--only_one_samplefolder', dest='only_one_samplefolder', action='store_true')
     parser.add_argument('--load_subnet', dest='load_subnet', action='store_true')
     parser.add_argument('--EarlyStop', dest='EarlyStop',  type=str, default='EarlyStop')
+    parser.add_argument('--max_idx', dest='max_idx',  type=int, default=-1)
+    parser.add_argument('--dense_img_num', dest='dense_img_num',  type=int, default=-1)
+
     args = parser.parse_args()
 
 parser()
@@ -85,6 +88,8 @@ def printdeb(*what):
 class OurMiniImageNetDataLoader(MiniImageNetDataLoader):
     # adding functions we need
     def idx_to_big(self, phase, idx):
+        if args.max_idx>0 and idx >= args.max_idx: #to limit the allowed different batches
+            return True
         if phase=='train':
             all_filenames = self.train_filenames
         elif phase=='val':
@@ -368,13 +373,16 @@ def get_config(self):
 conv4 = TimeDistributed(Conv2D(args.hidden_size, 3, padding='same', activation = 'relu', name = 'conv_4'))(pool3)
 pool4 = TimeDistributed(MaxPooling2D(pool_size = 2, name = 'pool_4'))(conv4)
 conv5 = TimeDistributed(Conv2D(args.hidden_size, 3, padding='same', activation = 'relu', name = 'conv_5'))(pool4)
-pool5 = TimeDistributed(MaxPooling2D(pool_size = 5, name = 'pool_5'))(conv5)
+pool5 = TimeDistributed(MaxPooling2D(pool_size = 2, name = 'pool_5'))(conv5)
 
 flat = TimeDistributed(Flatten())(pool5)
-#x = TimeDistributed(Dense(100, activation = 'relu'))(flat)
+if args.dense_img_num > 0:
+    x = TimeDistributed(Dense(args.dense_img_num, activation = 'tanh'))(flat)
+else:
+    x = flat
 #predictions = Activation('softmax')(x)
 
-model_img = FindModel(inputs=inputs, outputs=flat)
+model_img = FindModel(inputs=inputs, outputs=x)
 
 #model_img.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['categorical_accuracy'])
 
@@ -383,7 +391,7 @@ def get_config(self):
 input1 = Input(shape=(None,84,84,3))
 input2 = Input(shape=(None,84,84,3)) #, tensor = K.variable(episode_train_img[0:0]))
 
-input2b = BiasLayer(shots * cathegories, args.biaslayer1, bias_num = 1, name = 'bias1'+str(cathegories)+'_'+str(args.shots)+'t')(input2)
+input2b = BiasLayer(shots * cathegories, args.biaslayer1, bias_num = 1, name = 'bias1_'+str(cathegories)+'_'+str(args.shots)+'t')(input2)
 encoded_l = model_img(input1)
 encoded_r = model_img(input2b)