[FIXED] Tensorflow - matmul is showing InvalidArgumentError

Issue

I am trying to use matmul in tensorflow like this-

tf.matmul(y_true, logy)

where

y_true = <class 'numpy.ndarray'> (1000, 1, 1)

logy = <class 'tensorflow.python.framework.ops.EagerTensor'> (1000, 1)

I am getting this error-

<class 'numpy.ndarray'> (1000, 1, 1)
<class 'tensorflow.python.framework.ops.EagerTensor'> (1000, 1)
---------------------------------------------------------------------------
InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-39-4cc645f4bb75> in <module>()
     25   with tf.GradientTape() as tape:
     26     y_predicted = h(x_train,w)
---> 27     costF = loss_function(y_predicted, y)
     28 
     29   gradients = tape.gradient(costF, w)

5 frames
<ipython-input-39-4cc645f4bb75> in loss_function(y_pred, y_true)
     18   print(type(y_true), y_true.shape)
     19   print(type(logy), logy.shape)
---> 20   p1 = tf.matmul(y_true, logy)
     21   p2 = tf.matmul(1-y_true, logyInv)
     22   return -tf.reduce_mean(p1 + p2)

/usr/local/lib/python3.7/dist-packages/tensorflow/python/util/dispatch.py in wrapper(*args, **kwargs)
    204     """Call target, and fall back on dispatchers if there is a TypeError."""
    205     try:
--> 206       return target(*args, **kwargs)
    207     except (TypeError, ValueError):
    208       # Note: convert_to_eager_tensor currently raises a ValueError, not a

/usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/math_ops.py in matmul(a, b, transpose_a, transpose_b, adjoint_a, adjoint_b, a_is_sparse, b_is_sparse, output_type, name)
   3606       else:
   3607         return gen_math_ops.batch_mat_mul_v2(
-> 3608             a, b, adj_x=adjoint_a, adj_y=adjoint_b, name=name)
   3609 
   3610     # Neither matmul nor sparse_matmul support adjoint, so we conjugate

/usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/gen_math_ops.py in batch_mat_mul_v2(x, y, adj_x, adj_y, name)
   1507       return _result
   1508     except _core._NotOkStatusException as e:
-> 1509       _ops.raise_from_not_ok_status(e, name)
   1510     except _core._FallbackException:
   1511       pass

/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/ops.py in raise_from_not_ok_status(e, name)
   6939   message = e.message + (" name: " + name if name is not None else "")
   6940   # pylint: disable=protected-access
-> 6941   six.raise_from(core._status_to_exception(e.code, message), None)
   6942   # pylint: enable=protected-access
   6943 

/usr/local/lib/python3.7/dist-packages/six.py in raise_from(value, from_value)

InvalidArgumentError: cannot compute BatchMatMulV2 as input #1(zero-based) was expected to be a int64 tensor but is a double tensor [Op:BatchMatMulV2]

Can anyone please help?

Solution

As @Karl Knechtel already said, you have a mismatch in your shapes. Correct it and see if the error goes away.

If not, it's because you're trying to perform an operation with operands of different types (int64 and double). The message says what's wrong: InvalidArgumentError: cannot compute BatchMatMulV2 as input #1(zero-based) was expected to be a int64 tensor but is a double tensor [Op:BatchMatMulV2]. It says that your first argument is int64 and your second argument is double. Cast either of your arrays:

y_true.astype(np.double)

or

logy = tf.cast(logy, tf.int64)

Answered By - Ladislav Ondris