TypeError:s [p] .shape在Firefox中未定义

时间:2018-08-25 06:55:08

标签: javascript tensorflowjs-converter

 model = tf.sequential();

 a = tf.layers.input({shape: [127,]});
 b = tf.layers.dense({units:64, inputShape: [127,], activation: 'relu' }).apply(a)
 c = tf.layers.dropout(0.5).apply(b)
 d = tf.layers.dense({units:64, activation: 'relu'}).apply(c)
 e = tf.layers.dropout(0.5).apply(d)
 f = tf.layers.dense({units:12, activation: 'sigmoid'}).apply(e)

 model = tf.model({inputs: a, outputs: f});     

 model.compile({
    optimizer: 'rmsprop', 
    loss: 'meanSquaredError', 
    metrics: 'accuracy'
 });

 console.log(model.summary());

 await model.fit(input1:XS[1].split(','), {main_output:YS[1].split(',')}, {epochs: 50});

这里XS [1]的形状为(127,),YS [1]的形状为(12,),我根据'\ n'和','进行了分割,仍然存在错误。任何帮助将不胜感激!

1 个答案:

答案 0 :(得分:1)

我通过以某种方式包装输入和输出来修复它 {input1:tf.tensor(Array(XS [1] .split(','))))},{main_output:tf.tensor(Array(YS [1] .split(',')))}}