diff --git a/examples/language-model/make_data.py b/examples/language-model/make_data.py
index 6945cd4..ddf3c99 100644
--- a/examples/language-model/make_data.py
+++ b/examples/language-model/make_data.py
@@ -39,17 +39,17 @@ def one_hot(v, ndim):
     'pride_and_prejudice.txt',
     'shakespeare.txt'
 ]:
-    with open('books/%s' % book, 'r') as infile:
+    with open('books/%s' % book, 'r', encoding='utf-8') as infile:
         chars = [
             c for c in ' '.join(infile.read().lower().split())
-            if c in set(vocab)
+            if c in vocab
         ]
         all_chars += [' ']
         all_chars += chars
 
 all_chars = list(' '.join(''.join(all_chars).split()))
 num_chars = len(all_chars)
-with open('cleaned.txt', 'w') as outfile:
+with open('cleaned.txt', 'w', encoding='utf-8') as outfile:
     outfile.write(''.join(all_chars))
 
 
@@ -71,10 +71,7 @@ def one_hot(v, ndim):
             data_portions[i][1] * 0.1
         )
 
-max_i = sum([
-    int(round(len(all_chars) * fraction))
-    for name, fraction in data_portions
-]) - seq_len
+max_i = sum(int(round(len(all_chars) * fraction)) for (name, fraction) in data_portions) - seq_len
 
 for i in range(max_i):
 
@@ -105,7 +102,7 @@ def one_hot(v, ndim):
 
     start_i = end_i
 
-    with open('data/%s.jsonl' % name, 'w') as outfile:
+    with open('data/%s.jsonl' % name, 'w', encoding='utf-8') as outfile:
         for sample_x, sample_y in zip(x0, y0):
             outfile.write(json.dumps({
                 'in_seq': sample_x.tolist(),
@@ -113,4 +110,4 @@ def one_hot(v, ndim):
             }))
             outfile.write('\n')
 
-    del x0, y0
+    del x0, y0
\ No newline at end of file
diff --git a/examples/language-model/view_outputs.py b/examples/language-model/view_outputs.py
index 070af47..997bd43 100644
--- a/examples/language-model/view_outputs.py
+++ b/examples/language-model/view_outputs.py
@@ -50,13 +50,6 @@
     else:
         print((' ' * (seq_len + 5)) + 'INCORRECT (%s)' % correct_char)
 
-accuracy = sum(
-    [
-        int(
-            np.argmax(prediction_data['result']['out_char'][i]) == np.argmax(prediction_data['truth']['out_char'][i])
-        )
-        for i in range(batch_size)
-    ]
-) / float(len(prediction_data['truth']['out_char']))
+accuracy = sum(int(np.argmax(prediction_data['result']['out_char'][i]) == np.argmax(prediction_data['truth']['out_char'][i])) for i in range(batch_size)) / float(len(prediction_data['truth']['out_char']))
 
-print('accuracy = %s' % accuracy)
+print('accuracy = %s' % accuracy)
\ No newline at end of file
diff --git a/kur/__main__.py b/kur/__main__.py
index 4b471bf..a0fcbab 100644
--- a/kur/__main__.py
+++ b/kur/__main__.py
@@ -24,9 +24,9 @@
 import logging
 import kur.__main__
 
-from . import __version__, __homepage__
+from . import Kurfile, __homepage__, __version__
 from .utils import logcolor
-from . import Kurfile
+
 from .plugins import Plugin
 from .engine import JinjaEngine
 
@@ -87,10 +87,7 @@ def build(args):
 	spec = parse_kurfile(args.kurfile, args.engine)
 
 	if args.compile == 'auto':
-		result = []
-		for section in ('train', 'test', 'evaluate'):
-			if section in spec.data:
-				result.append((section, 'data' in spec.data[section]))
+		result = [(section, 'data' in spec.data[section]) for section in ('train', 'test', 'evaluate') if section in spec.data]
 		if not result:
 			logger.info('Trying to build a bare model.')
 			args.compile = 'none'
@@ -118,7 +115,7 @@ def build(args):
 
 	if args.compile == 'none':
 		return
-	elif args.compile == 'train':
+	if args.compile == 'train':
 		target = spec.get_trainer(with_optimizer=True)
 	elif args.compile == 'test':
 		target = spec.get_trainer(with_optimizer=False)
@@ -452,7 +449,7 @@ def main():
 		if gotcha:
 			plugin_dir = arg
 			break
-		elif arg == '--plugin':
+		if arg == '--plugin':
 			gotcha = True
 	plugin_dir = plugin_dir or os.environ.get('KUR_PLUGIN')
 	load_plugins(plugin_dir)