Question

我正在使用以tensorflow编写的wavenet进行语音到文本的转换。通过烧瓶部署模型。运行烧瓶模型时，模型第一次运行良好。但是当我重新发送请求时，它显示了tensorflow图错误。 PS：如果我停下烧瓶，然后再次运行模型，则对于第一个请求，它工作正常。但是当我再次发送请求时，它再次显示该错误。

@app.route('/',methods=['GET','POST'])
def transcribe():

# video: list of base64 of frames
if request.method == 'POST':              
    from model import get_logit
    import sugartensor as tf
    # set log level to debug

    tf.sg_verbosity(10)

    num_blocks = 3     # dilated blocks
    num_dim = 128      # latent dimension

    import data
    import os
    os.environ['TF_CPP_MIN_LOG_LEVEL']='2'

    batch_size = 1     # batch size
    voca_size = data.voca_size
    print(voca_size)

    # mfcc feature of audio
    x = tf.placeholder(dtype=tf.sg_floatx, shape=(batch_size, None, 20))
    # sequence length except zero-padding
    seq_len = tf.not_equal(x.sg_sum(axis=2), 0.).sg_int().sg_sum(axis=1)

    # encode audio feature
    logit = get_logit(x, voca_size=voca_size)
    
    # ctc decoding
    decoded, _ = tf.nn.ctc_beam_search_decoder(logit.sg_transpose(perm=[1, 0, 2]), seq_len, merge_repeated=False)

    # to dense tensor
    y = tf.sparse_to_dense(decoded[0].indices, decoded[0].dense_shape, decoded[0].values) + 1

    #
    # regcognize wave file
    #
    download_path= "/home/user/Downloads/audio.flac"
    shutil.move(download_path, os.getcwd()+ '/static/audio/audio.flac')
    # command line argument for input wave file path
    tf.sg_arg_def(file=(os.getcwd()+ '/static/audio/audio.flac', 'speech wave file to recognize.'))

    # load wave file
    wav, _ = librosa.load(tf.sg_arg().file, mono=True, sr=16000)
    # get mfcc feature
    mfcc = np.transpose(np.expand_dims(librosa.feature.mfcc(wav, 16000), axis=0), [0, 2, 1])

    # run network
    with tf.Session() as sess:

        # init variables
        tf.sg_init(sess)

        # restore parameters
        saver = tf.train.Saver()
        saver.restore(sess, tf.train.latest_checkpoint('static/asset/train'))
        # run session
        label = sess.run(y, feed_dict={x: mfcc})

        # print label
        sentence= data.print_index(label)
        return render_template("index.html",transcribed=sentence)

它显示的错误是这样的： ValueError：Tensor（“ cond / pred_id：0”，shape =（），dtype = bool）必须与Tensor（“ front / conv_in / moments / normalize / mean：0”，shape =（128，）来自同一张图，dtype = float32）。

它显示logit = get_logit（x，voca_size = voca_size）中的错误在get_logit中在此框架中打开一个交互式python shellz = x.sg_conv1d（size = 1，dim = num_dim，act ='tanh'，bn = True，name ='conv_in'）

Tensorflow图显示烧瓶重新加载错误

0 个答案: