尝试使用音乐数据进行迁移学习

Question

我正在尝试在以下存储库中执行 github 代码：

https://github.com/shubham3121/music-generation-using-rnn 错误在以下函数中：

def generate():
    """ Generate a piano midi file """
    #load the notes used to train the model
    with open('data/notes', 'rb') as filepath:
        notes = pickle.load(filepath)

    # Get all pitch names
    pitchnames = sorted(set(item for item in notes))
    # Get all pitch names
    n_vocab = len(set(notes))

    print('Initiating music generation process.......')

    network_input = get_inputSequences(notes, pitchnames, n_vocab)
    model = create_network(normalized_input, n_vocab)
    print('Loading Model weights.....')
    model.load_weights('weights.best.music3.hdf5')
    print('Model Loaded')
    prediction_output = generate_notes(model, network_input, pitchnames, n_vocab)
    create_midi(prediction_output)

问题如下：

Initiating music generation process.......

---------------------------------------------------------------------------

NameError                                 Traceback (most recent call last)

<ipython-input-24-b97e0091fa1b> in <module>()
----> 1 generate()

<ipython-input-20-27d37a8858b8> in generate()
     13 
     14     network_input = get_inputSequences(notes, pitchnames, n_vocab)
---> 15     model = create_network(normalized_input, n_vocab)
     16     print('Loading Model weights.....')
     17     model.load_weights('weights.best.music3.hdf5')

NameError: name 'normalized_input' is not defined

由于代码不是我的，我不知道这个变量是什么，所以我可以修复它... 我给代码所有者发了电子邮件，但他没有回复。

谢谢

Answer 1

我假设实际代码是here

不是 100% 肯定，但很可能是代码中的错误。应该是 network_input。所以你改变了

network_input = get_inputSequences(notes, pitchnames, n_vocab)
model = create_network(normalized_input, n_vocab)

到

network_input = get_inputSequences(notes, pitchnames, n_vocab)
model = create_network(network_input, n_vocab)

这会产生以下错误：

Initiating music generation process.......

---------------------------------------------------------------------------

AttributeError                            Traceback (most recent call last)

<ipython-input-27-b97e0091fa1b> in <module>()
----> 1 generate()

1 frames

<ipython-input-14-866d26da5fd7> in create_network(network_in, n_vocab)
      4     """Create the model architecture"""
      5     model = Sequential()
----> 6     model.add(LSTM(128, input_shape=network_in.shape[1:], return_sequences=True))
      7     model.add(Dropout(0.2))
      8     model.add(LSTM(128, return_sequences=True))

AttributeError: 'list' object has no attribute 'shape'

好的，如果你看一下来自

的代码

In [7]: def prepare_sequences(notes, n_vocab):

你可以看到转换为 numpy 数组：

network_input = np.reshape(network_input, (n_patterns, sequence_length, 1))

您应该执行类似于 get_inputSequences 的操作，以便返回 numpy 数组，而不是常规的 python 列表。但是代码看起来很旧，所以即使你修复了错误，你也可能会得到意想不到的结果，所以最好在 github 下弄清楚。或者，您需要修复并确保获得良好的结果，否则您需要参与代码。