- was necessary to add changes to mini_imagenet_dataloader.py as we n…

…eeded unsampled first training set to save model_img output to BiasLayer 2

- output saving seems to be OK

few_shot_tests.py

@@ -37,6 +37,7 @@
 parser.add_argument('--biaslayer2', dest='biaslayer2', action='store_true')
 parser.add_argument('--shots', dest='shots',  type=int, default=5)
 parser.add_argument('--debug', dest='debug', action='store_true')
+parser.add_argument('--set_model_img_to_weights', dest='set_model_img_to_weights', action='store_true')
 
 args = parser.parse_args()
 
@@ -110,7 +111,7 @@ def generate_add_samples(self, phase = 'train'):
         self.idx = 0
         while True:
             batch_train_img, batch_train_label, episode_test_img, episode_test_label = \
-                dataloader.get_batch(phase=args.dataset, idx=self.idx)
+                dataloader.get_batch(phase=args.dataset, idx=self.idx, dont_shuffle_batch = (self.idx==0))
 
             # this depends on what we are trying to train.
             # care must be taken, that with a different dataset the labels have a different meaning. Thus if we use a new dataset, we must 
@@ -227,7 +228,7 @@ def build(self, input_shape):
                                       shape=(1),
                                       initializer=preset,
                                       trainable=False)
-        #print('bias_enable',self.bias_enable, K.eval(self.bias_enable[0]),'bias',self.bias,'weights')
+        print('bias_enable',self.bias_enable, K.eval(self.bias_enable[0]),'bias',self.bias,'weights')
         super(BiasLayer, self).build(input_shape)  # Be sure to call this at the end
 
     def set_bias(self, do_bias):
@@ -320,7 +321,7 @@ def call(x):
 
 if args.pretrained_name is not None:
     from tensorflow.keras.models import load_model
-    lambda_model = load_model(args.pretrained_name, custom_objects = { "keras": tensorflow.keras , "args":args, "BiasLayer": BiasLayer})
+    lambda_model = load_model(args.pretrained_name, custom_objects = { "keras": tensorflow.keras , "args":args, "BiasLayer": BiasLayer, "FindModel": FindModel})
     print("loaded model",lambda_model)
 
 # models in models forget the layer name, therefore one must use the automatically given layer name and iterate throught the models by hand
@@ -383,6 +384,7 @@ def all_layers(model):
 #print('test lambda', K.eval(test_lambda))
 
 
+print('vor fitting', lambda_model_layers[17].get_weights()[0])
 
 checkpointer = ModelCheckpoint(filepath='checkpoints/model-{epoch:02d}.hdf5', verbose=1)
 tensorboard = TensorBoard(log_dir = args.tensorboard_logdir)
@@ -426,9 +428,18 @@ def print_FindModels(model):
 out_test = find_conv_model.output
 functor = K.function([in_test], [out_test])
 
-calc_out = functor([K.expand_dims(K.expand_dims(K.variable(keras_gen_train.n_b_i),axis=0),axis=0)])
+calc_out = functor([K.expand_dims(K.variable(keras_gen_train.n_b_i),axis=0)])
+
+print('calc_out',calc_out[0])
+
+print('vor', lambda_model_layers[17].get_weights()[0])
+
+if args.set_model_img_to_weights:
+    print('\n!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!')
+    lambda_model_layers[17].set_weights([calc_out[0][0],np.array([0])])
+    print('nach', lambda_model_layers[17].get_weights()[0])
+    print('!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n')
 
-print(calc_out)
 
 
 
@@ -447,7 +458,7 @@ def print_FindModels(model):
             l2.do_bias = l2.trainable = args.biaslayer1
         if (l2.bias_num == 2):
             l2.do_bias = l2.trainable = args.biaslayer2
-        print('past',l2.bias_num, l2.do_bias,args.biaslayer1,args.biaslayer2, l2.bias)
+        print('past',l2.bias_num, l2.do_bias,args.biaslayer1,args.biaslayer2, debug(l2.bias))
 
     print('{:10} {:10} {:20} {:10}  {:10}'.format(l, p,l2.name, ("fixed", "trainable")[l2.trainable], l2.count_params()), debug(l2.get_weights()))
 

mini_imagenet_dataloader.py

@@ -141,12 +141,13 @@ def load_list(self, phase='train'):
         else:
             print('Please select vaild phase')
 
-    def process_batch(self, input_filename_list, input_label_list, batch_sample_num, reshape_with_one=True):
+    def process_batch(self, input_filename_list, input_label_list, batch_sample_num, reshape_with_one=True, dont_shuffle_batch = False):
         new_path_list = []
         new_label_list = []
         for k in range(batch_sample_num):
             class_idxs = list(range(0, self.way_num))
-            random.shuffle(class_idxs)
+            if not dont_shuffle_batch:
+                random.shuffle(class_idxs)
             for class_idx in class_idxs:
                 true_idx = class_idx*batch_sample_num + k
                 new_path_list.append(input_filename_list[true_idx])
@@ -174,7 +175,7 @@ def one_hot(self, inp):
             out[idx, inp[idx]] = 1
         return out
 
-    def get_batch(self, phase='train', idx=0):
+    def get_batch(self, phase='train', idx=0, dont_shuffle_batch = False):
         if phase=='train':
             all_filenames = self.train_filenames
             labels = self.train_labels 
@@ -204,7 +205,7 @@ def get_batch(self, phase='train', idx=0):
             this_task_te_filenames += this_class_filenames[epitr_sample_num:]
             this_task_te_labels += this_class_label[epitr_sample_num:]
 
-        this_inputa, this_labela = self.process_batch(this_task_tr_filenames, this_task_tr_labels, epitr_sample_num, reshape_with_one=False)
-        this_inputb, this_labelb = self.process_batch(this_task_te_filenames, this_task_te_labels, epite_sample_num, reshape_with_one=False)
+        this_inputa, this_labela = self.process_batch(this_task_tr_filenames, this_task_tr_labels, epitr_sample_num, reshape_with_one=False, dont_shuffle_batch = dont_shuffle_batch)
+        this_inputb, this_labelb = self.process_batch(this_task_te_filenames, this_task_te_labels, epite_sample_num, reshape_with_one=False, dont_shuffle_batch = dont_shuffle_batch)
 
         return this_inputa, this_labela, this_inputb, this_labelb