WARNING:tensorflow:Layers in a Sequential model should only have a single input tensor

Question

I have copy past code from tensorflow website's introduction to autoencoder first examplefollowing code works with mnist fashion dataset but not mine.This gives me a very long warning.Please tell me what is worng with my dataset the warning screen short of same error

here x_train is my dataset:

tf.shape(x_train)

output <tf.Tensor: shape=(3,), dtype=int32, numpy=array([169,**  **28,  28])>

here x_train is the mnist dataset:

tf.shape(x_train)

output<tf.Tensor: shape=(3,), dtype=int32, numpy=array([60000,    28,    28])>

My whole code to make dataset:

dir_path='auto/ttt/'
data=[]
x_train=[]
for i in os.listdir(dir_path):
    img=image.load_img(dir_path+'//'+i,color_mode='grayscale',target_size=(28,28))  
    data=np.array(img)
    data=data/255.0
    x_train.append(data)

this is the warning:

WARNING:tensorflow:Layers in a Sequential model should only have a single input tensor. Received: inputs=(<tf.Tensor 'IteratorGetNext:0' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:1' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:2' shape=(None, 28) 
dtype=float32>, <tf.Tensor 'IteratorGetNext:3' shape=(None, 28) 
dtype=float32>, <tf.Tensor 'IteratorGetNext:4' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:5' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:6' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:7' shape=(None, 28) dtype=flo...

also this value error (same warning):

ValueError: Exception encountered when calling layer "sequential_4" (type Sequential).
        
Layer "flatten_2" expects 1 input(s), but it received 169 input tensors. Inputs received: [<tf.Tensor 'IteratorGetNext:0' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:1' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:2' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:3' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:4' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:5' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:6' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:7' shape=(None, 28) dtype=float32>, <tf.Tensor 'IteratorGetNext:8' shape=(None, 28) dtype=float3...

Answer 1

The model.fit() is given a list of arrays as input. A list of arrays is generally passed to fit() when a model has multiple inputs. In this case, the fit() method is treating each array as an input, resulting in the error.

Please convert the data to a tensor as follows and try again.

x_train=tf.convert_to_tensor(x_train) 

Refer to the gist for complete code.