ValueError: Shape must be rank 1 but is rank 0 for 'random_normal_6/RandomStandardNormal' (op: 'RandomStandardNormal') with input shapes: [].

[Rajapradeepan]

getting this error for the code
n_input = 784
n_hidden_1=256
weights={'h1': tf.global_variables_initializer(tf.random_normal(n_input ,n_hidden_1)),
'h2': tf.global_variables_initializer(tf.random_normal(n_hidden_1,n_hidden_2)),
'out': tf.global_variables_initializer(tf.random_normal(n_hidden_2,n_classes))
}

[Pedram Hosseini]

instead of
tf.random_normal(n_input ,n_hidden_1)
use the following (add brackets):
tf.random_normal([n_input ,n_hidden_1])

[Victor]

If I wrote them separated to a single variable ,instead of using dictionary, it won't work and threw an compiling error, like the snippet code below:

n_input = 784         
n_classes = 10        
n_hidden_layer = 256

hidden_layer_weight = tf.Variable(tf.random_normal([n_input, n_hidden_layer]))
hidden_output = tf.Variable(tf.random_normal([n_hidden_layer, n_classes]))

So , this one could be working.

weights = {
    'hidden_layer': tf.Variable(tf.random_normal([n_input, n_hidden_layer])),
    'out': tf.Variable(tf.random_normal([n_hidden_layer, n_classes]))
}