TensorFlowのsaver.restoreについて

Python3のTensorFlowにて、下記のコードによりエラーが発生しました。
エラーの原因はtf.train.Saverにあるようなのですが、調べたところTensorFlowのバージョンが悪いのやら、
cifar10.pyの一部を書き換えが必要なのやら、何が正しいのかわからない状態です。

どなたかわかる方がいらっしゃれば、お教えいただけないでしょうか。

Python3
1import sys
2import tensorflow as tf
3import cifar10
4
5FLAGS = tf.app.flags.FLAGS
6cifar10.NUM_CLASSES = 7
7tf.app.flags.DEFINE_string('checkpoint_dir', '/tmp/cifar10_train',
8                           """Directory where to read model checkpoints.""")
9
10def evaluate(filename):
11  with tf.Graph().as_default() as g:
12    file   = tf.read_file(filename)
13    image  = tf.image.decode_jpeg(file, channels = 3)
14    image  = tf.image.resize_images(image, [32, 32])
15    # cifar10は内部処理で32×32を24×24に切り出して利用している
16    image  = tf.image.resize_image_with_crop_or_pad(image, 24, 24)
17    logits = cifar10.inference([image])
18
19    # ここのkの値もクラス数と一致させる
20    top_k_op = tf.nn.top_k(logits, k=7)
21
22    variable_averages = tf.train.ExponentialMovingAverage(cifar10.MOVING_AVERAGE_DECAY)
23    variables_to_restore = variable_averages.variables_to_restore()
24    saver = tf.train.Saver(variables_to_restore)
25
26    sess = tf.InteractiveSession()
27    sess.run(tf.initialize_all_variables())
28    ckpt = tf.train.get_checkpoint_state(FLAGS.checkpoint_dir)
29
30    if ckpt and ckpt.model_checkpoint_path:
31      saver.restore(sess, ckpt.model_checkpoint_path)
32
33    else:
34      print('No checkpoint file found')
35      return
36
37    tf.train.start_queue_runners(sess=sess)
38    values, indices = sess.run(top_k_op)
39    ratio = sess.run(tf.nn.softmax(values[0]))
40
41    # 予想したラベルとそれぞれに対する確信度
42    print(indices[0])
43    print(ratio)
44
45def main(argv=None):
46  evaluate(sys.argv[1])
47
48if __name__ == '__main__':
49  tf.app.run()

エラー内容は下記の通りです。

Python3
1Traceback (most recent call last):
2  File "cifar10_sample.py", line 60, in <module>
3    tf.app.run()
4  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/platform/app.py", line 44, in run
5    _sys.exit(main(_sys.argv[:1] + flags_passthrough))
6  File "cifar10_sample.py", line 56, in main
7    evaluate(sys.argv[1])
8  File "cifar10_sample.py", line 40, in evaluate
9    saver.restore(sess, ckpt.model_checkpoint_path)
10  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/training/saver.py", line 1439, in restore
11    {self.saver_def.filename_tensor_name: save_path})
12  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 767, in run
13    run_metadata_ptr)
14  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 965, in _run
15    feed_dict_string, options, run_metadata)
16  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1015, in _do_run
17    target_list, options, run_metadata)
18  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1035, in _do_call
19    raise type(e)(node_def, op, message)
20tensorflow.python.framework.errors_impl.InvalidArgumentError: Assign requires shapes of both tensors to match. lhs shape= [192,7] rhs shape= [192,10]
21	 [[Node: save/Assign_9 = Assign[T=DT_FLOAT, _class=["loc:@softmax_linear/weights"], use_locking=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/cpu:0"](softmax_linear/weights, save/RestoreV2_9)]]
22
23Caused by op 'save/Assign_9', defined at:
24  File "cifar10_sample.py", line 60, in <module>
25    tf.app.run()
26  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/platform/app.py", line 44, in run
27    _sys.exit(main(_sys.argv[:1] + flags_passthrough))
28  File "cifar10_sample.py", line 56, in main
29    evaluate(sys.argv[1])
30  File "cifar10_sample.py", line 33, in evaluate
31    saver = tf.train.Saver(variables_to_restore)
32  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/training/saver.py", line 1051, in __init__
33    self.build()
34  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/training/saver.py", line 1081, in build
35    restore_sequentially=self._restore_sequentially)
36  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/training/saver.py", line 675, in build
37    restore_sequentially, reshape)
38  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/training/saver.py", line 414, in _AddRestoreOps
39    assign_ops.append(saveable.restore(tensors, shapes))
40  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/training/saver.py", line 155, in restore
41    self.op.get_shape().is_fully_defined())
42  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/ops/gen_state_ops.py", line 47, in assign
43    use_locking=use_locking, name=name)
44  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/framework/op_def_library.py", line 763, in apply_op
45    op_def=op_def)
46  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 2395, in create_op
47    original_op=self._default_original_op, op_def=op_def)
48  File "/Users/k_aki86/.pyenv/versions/3.5.1/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 1264, in __init__
49    self._traceback = _extract_stack()
50
51InvalidArgumentError (see above for traceback): Assign requires shapes of both tensors to match. lhs shape= [192,7] rhs shape= [192,10]
52	 [[Node: save/Assign_9 = Assign[T=DT_FLOAT, _class=["loc:@softmax_linear/weights"], use_locking=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/cpu:0"](softmax_linear/weights, save/RestoreV2_9)]]