Get list of object in tensorflow iterating tensor-CodePudding

So I wanna return a list of object in tensorflow, the list object is constructed using tensor X with size [n,k,d] when i tried:

listofobject = [Myobject(x) for x in X]

it returns:

OperatorNotAllowedInGraphError: iterating over `tf.Tensor` is not allowed: AutoGraph did convert this function. This might indicate you are trying to use an unsupported feature.

I also tried:

listofobject = tf.map_fn(lambda x:Myobject(x), X)

and got this error:

 ValueError: Tried to convert 'y' to a tensor and failed. Error: None values not supported.

when I check inside the object's constructor, at some point my x shape is [None, d] not sure what I am missing here. My question is, what is the proper way to return a list of objects like this in tensorflow?

CodePudding user response：

This works for me,

import tensorflow as tf

class Myobject:

    def __init__(self, x):
       self.x = x


X = tf.random.uniform(shape = (3, 2 ,1))

listofobject = [Myobject(x) for x in X]

print(listofobject[1].x)
<tf.Tensor: shape=(2, 1), dtype=float32, numpy=
array([[0.09055722],
       [0.7195903 ]], dtype=float32)>

CodePudding user response：

Could you display more detail on why you need it to.

Two Methods:

tf.ragged.constant([Myobject(x) for x in X])

tf.map_fn(Myobject,elems=X,fn_ouput_signature=tf.RaggedTensorSpec(dtype=tf.float32orWHATEVER))

Because it is ragged that is the problem or the values are placeholders could you please alboarate.