Learn to normalize characters given a line, and using the model of the no-learning version

36-dnn/normalize-chars.py

@@ -1,5 +1,5 @@
 from keras.models import Model
-from keras import layers
+from keras import layers, metrics
 from keras.layers import Input, Dense
 from keras.utils import plot_model
 
@@ -39,18 +39,16 @@ def decode_one_hot(x):
     """
     s = []
     for onehot in x:
-        one_index = np.where(onehot == 1) # one_index is a tuple of two things
-        if len(one_index[0]) > 0:
-            n = one_index[0][0]
-            c = indices_char[n]
-            s.append(c) 
+        one_index = np.argmax(onehot)
+        c = indices_char[one_index]
+        s.append(c) 
     return ''.join(s)
-    
+
 def build_model():
     print('Build model...')
     
     # Normalize every character in the input, using a shared dense model
-    n_layer = Dense(INPUT_VOCAB_SIZE)
+    n_layer = Dense(INPUT_VOCAB_SIZE, activation = "softmax")
     raw_inputs = []
     normalized_outputs = []
     for _ in range(0, LINE_SIZE):
@@ -101,7 +99,7 @@ plot_model(model, to_file='normalization.png', show_shapes=True)
 # Train the model each generation and show predictions against the validation
 # dataset.
 val_gen2 = input_generator(1)
-for iteration in range(1, 500):
+for iteration in range(1, 12):
     print()
     print('-' * 50)
     print('Iteration', iteration)
@@ -112,40 +110,33 @@ for iteration in range(1, 500):
                 steps_per_epoch = 20,
                 validation_data = val_gen,
                 validation_steps = 10, workers=1)
-    # Select 10 samples from the validation set at random so we can visualize
-    # errors.
-#    print(batch_y)
-#    print(preds)
+    # Select samples from the a set at random so we can visualize errors.
     batch_x, batch_y = next(val_gen2)
     for i in range(len(batch_y)):
         preds = model.predict(batch_x)
         expected = batch_y[i]
         prediction = preds[i]
-        #print(preds)
-#        preds[preds>=0.5] = 1
-#        preds[preds<0.5] = 0
 
-        #q = ctable.decode(query)
         correct = decode_one_hot(expected)
         guess = decode_one_hot(prediction)
-        print('T', correct)
-        print('G', guess)
+        print('T:', correct)
+        print('G:', guess)
 
-#with open(sys.argv[1]) as f:
-#    for line in f:
-#        if line.isspace(): continue
-#        onehots = encode_one_hot(line)
+with open(sys.argv[1]) as f:
+    for line in f:
+        if line.isspace(): continue
+        onehots = encode_one_hot(line)
 
-#        data = [[] for _ in range(LINE_SIZE)]
-#        for i, c in enumerate(onehots):
-#            data[i].append(c)
-#        for j in range(len(onehots), LINE_SIZE):
-#            data[j].append(np.zeros((INPUT_VOCAB_SIZE)))
+        data = [[] for _ in range(LINE_SIZE)]
+        for i, c in enumerate(onehots):
+            data[i].append(c)
+        for j in range(len(onehots), LINE_SIZE):
+            data[j].append(np.zeros((INPUT_VOCAB_SIZE)))
 
-#        inputs = [np.array(e) for e in data]
+        inputs = [np.array(e) for e in data]
 
-#        preds = model.predict(inputs)
-#        normal = decode_one_hot(preds[0])
+        preds = model.predict(inputs)
+        normal = decode_one_hot(preds[0])
 
-#        print(decode_one_hot(onehots))
-#        print(normal)
+        print(decode_one_hot(onehots))
+        print(normal)