unhashable type: 'numpy.ndarray' error in tensorflow

Question

data = pd.read_excel("/Users/madhavthaker/Downloads/Reduced_Car_Data.xlsx")

train = np.random.rand(len(data)) < 0.8

data_train = data[train]
data_test = data[~train]


x_train = data_train.ix[:,0:3].values
y_train = data_train.ix[:,-1].values
x_test = data_test.ix[:,0:3].values
y_test = data_test.ix[:,-1].values

y_label = tf.placeholder(shape=[None,1], dtype=tf.float32, name='y_label')
x = tf.placeholder(shape=[None,3], dtype=tf.float32, name='x')
W = tf.Variable(tf.random_normal([3,1]), name='weights')
b = tf.Variable(tf.random_normal([1]), name='bias')
y = tf.matmul(x,W)  + b

init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init)
    summary_op = tf.summary.merge_all()
    #Fit all training data
    for epoch in range(1000):
        sess.run(train, feed_dict={x: x_train, y_label: y_train})

        # Display logs per epoch step
        if (epoch+1) % display_step == 0:
            c = sess.run(loss, feed_dict={x: x_train, y_label:y_train})
            print("Epoch:", '%04d' % (epoch+1), "cost=", "{:.9f}".format(c), \
                "W=", sess.run(W), "b=", sess.run(b))

    print("Optimization Finished!")
    training_cost = sess.run(loss, feed_dict={x: x_train, y_label: y_train})
    print("Training cost=", training_cost, "W=", sess.run(W), "b=", sess.run(b), '\n')

Here is the error:

x---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-37-50102cbac823> in <module>()
      6     #Fit all training data
      7     for epoch in range(1000):
----> 8         sess.run(train, feed_dict={x: x_train, y_label: y_train})
      9 
     10         # Display logs per epoch step

TypeError: unhashable type: 'numpy.ndarray'

Here are the shapes of both of the numpy arrays that I am inputting:

y_train.shape = (78,)
x_train.shape = (78, 3)

I have no idea what is causing this. All of my shapes match up and I shouldn't have any issues. Let me know if you need any more information.

Edit: From my comment on one of the answers below, it seems as though I had to specify a specific size for my placeholders. None was not satisfactory. When I changed that and re-ran my code, everything worked fine. Still not quite sure why that is.

Answer 1

In my case, the problem was naming the input parameter the same as the placeholder variable. This, of course, replaces your tensorflow variable with the input variable; resulting in a different key for the feed_dict.

A tensorflow variable is hashable, but your input parameter (np.ndarray) isn't. The unhashable error is therefore a result of you trying to pass your parameter as the key instead of a tensorflow variable. Some code to visualize what I'm trying to say:

a = tf.placeholder(dtype=tf.float32, shape=[1,2,3])
b = tf.identity(a)

with tf.Session() as sess:
    your_var = np.ones((1,2,3))
    a = your_var
    sess.run(b, feed_dict={a: a})

Hopes this helps anyone stumbling upon this problem in the future!

Answer 2

Please carefully check the datatype you feed "x_train/y_train" and the tensor "x/y_label" you defined by 'tf.placeholder(...)'

I have met the same problem with you. And the reason is x_train in my code is "np.float64", but what I defined by tf.placeholder() is tf.float32. The date type float64 and float32 is mismatching.