Question

尝试在Keras模型中运行光束搜索，我得到了混乱（和冲突？）错误消息。我的模型输入了诸如

inputs = Input(name='spectrograms',
               shape=(None, hparams["n_spectrogram"]))
input_length = Input(name='len_spectrograms',
                     shape=[1], dtype='int64')

和CTC损失函数要求输入和标签长度为[1]形状。据我了解，输出应该使用类似的

# Stick connectionist temporal classification on the end of the core model
paths = K.function(
    [inputs, input_length],
    K.ctc_decode(output, input_length, greedy=False, top_paths=4)[0])

但按原样，这导致对input_length

形状的抱怨

ValueError: Shape must be rank 1 but is rank 2 for 'CTCBeamSearchDecoder' (op: 'CTCBeamSearchDecoder') with input shapes: [?,?,44], [?,1].

但是如果我砍掉那个尺寸

    K.ctc_decode(output, input_length[..., 0], greedy=False, top_paths=4)[0])

模型定义运行，但是当我用y = paths([x, numpy.array([[30], [30]])])运行x.shape == (2, 30, 513)时，我突然得到了

tensorflow.python.framework.errors_impl.InvalidArgumentError: sequence_length is not a vector
     [[{{node CTCBeamSearchDecoder}} = CTCBeamSearchDecoder[beam_width=100, merge_repeated=true, top_paths=4, _device="/job:localhost/replica:0/task:0/device:CPU:0"](Log, ToInt32)]]

我在做什么错了？

CTCBeamSearchDecoder认为形状（2，）的sequence_length不是向量

0 个答案: