Failed to convert a NumPy array to a Tensor (Unsupported object type numpy.ndarray) error

Question

I try to train a tensorflow model. But I got error.

Failed to convert a NumPy array to a Tensor (Unsupported object type numpy.ndarray).

Here my fit codes:

model.fit(self.datas.trainImages, self.datas.trainLabels,self.datas.batch_size, epochs =self.datas.epochs)

My self.datas.trainImages is numpy.array() its shape is (16,) it has 16 sample and their sizes is 28x28, it is mnist dataset.

self.train_dataset = [[cv2.imread(image0),0],[cv2.imread(image1),1],[cv2.imread(image2),2],[...],[...]]
self.trainDataset = numpy.array(self.train_dataset)
        
self.trainImages, self.trainLabels = numpy.asarray(self.trainDataset[:,0])/255,self.trainDataset[:,1] #.astype(numpy.float32)/



self.val_dataset = [[cv2.imread(image0),0],[cv2.imread(image1),1],[cv2.imread(image2),2],[...],[...]]

self.valDataset = numpy.array(self.val_dataset)#.astype(numpy.float32)
self.valImages, self.valLabels = numpy.asarray(self.valDataset[:,0])/255,self.valDataset[:,1] #.astype(numpy.float32)/255

I tried to use astype or numpy.ndarray but I got another errors. I am sure of that all datas in the self.datas.trainImages is float numbers and has same shape.

What were the "another errors". What's the trainImages.dtype? — hpaulj
– hpaulj, Commented Dec 3, 2021 at 19:36
@hpaulj if I convert dtype into numpy.float 32, I get this error : setting an array element with a sequence. — Wtow
– Wtow, Commented Dec 3, 2021 at 20:22
That tells me that the array is object dtype, with elements that vary in shape. It can't be converted to a multidimensional numeric dtype array. Which is the root of the tensor conversion error. — hpaulj
– hpaulj, Commented Dec 3, 2021 at 20:33
@hpaulj I checked the element shapes in the self.trainImages, and all of them have same size (28,28) — Wtow
– Wtow, Commented Dec 4, 2021 at 6:56

Alex Knapp · Accepted Answer · 2021-12-04 15:18:55Z

2

Would it be possible to print out some type of error output?

Personally, I was having a similar issue and by coating my input with "np.stack()" it added an extra dimension, changed the shape of the array and allowed it to work.

i.e.

images = np.stack(self.data.trainImages)

EDIT: Removed irrelevant information

edited Dec 4, 2021 at 15:18

answered Dec 4, 2021 at 2:43

Alex Knapp

863 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Wtow Over a year ago

Thanks for your answer, your first answer is solved my problem. But it is weird because I used my same method before maybe 4-6 months and it was worked and I have not updated my tensorflow version. I got another error, I will open new post for new error.

Alex Knapp Over a year ago

Sorry, I mis-read the fit() call you did, the second part of my answer was for 2 inputs 1 output.

Wtow Over a year ago

Not problem, I got what you mean, thanks again everything.

Collectives™ on Stack Overflow

Failed to convert a NumPy array to a Tensor (Unsupported object type numpy.ndarray) error

1 Answer 1

3 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related