首先给出正确的调用:将InputDesc修改为:tf.TensorSpec。
以下是解释,可以选择跳过
今天在学习tensorfPack模块的时候,发现调用到tensorpack库里的InputDesc库的时候,发现这个问题:
from tensorpack.graph_builder.model_desc import ModelDesc, InputDesc
进而找到InputDesc库在代码中的使用如下:
return [InputDesc(tf.float32, (None, None, hp.default.n_mfcc), 'x_mfccs'),
InputDesc(tf.int32, (None, None,), 'y_ppgs')]
return [InputDesc(tf.float32, (None, n_timesteps, hp.default.n_mfcc), 'x_mfccs'),
InputDesc(tf.float32, (None, n_timesteps, hp.default.n_fft // 2 + 1), 'y_spec'),
InputDesc(tf.float32, (None, n_timesteps, hp.default.n_mels), 'y_mel'), ]
从而联想到InputDesc应该是给后面的变量赋值使用类型的,类似于TensorFlow里的tf.placeholder函数,构建一个占位符,但是查看tensorpack.graph_builder.model_desc原函数的时候,并没有发现InputDesc相关的函数,只有ModelDesc函数,上网查找也找不到,下班后也进行了相关的搜索,终于,功夫不负有心人,在万能网github上找到了相关解释:
https://github.com/tensorpack/tensorpack/blob/master/CHANGES.md
这里给出了详细解释:
2019/03/20. The concept of InputDesc was replaced by its equivalent in TF: tf.TensorSpec. This may be a breaking change if you have customized code that relies on internals of InputDesc. To use tf.TensorSpec in your ModelDesc:
def inputs(self):
return [tf.TensorSpec((None, 28, 28, 1), tf.float32, 'image'),
tf.TensorSpec((None,), tf.int32, 'label')]
符合“占位符”的理念。
至此大功告成,特记录一下!!!