K-best detector returns TypeError

[albusrice]

When running k-best = 4 and antenna configuration 128 x 128 BPSK i get the following error.

TypeError: Expected tensor [1. 2. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4.
 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4.
 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4.
 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4.
 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4. 4.
 4. 4. 4. 4. 4. 4. 4. 4. 4.] with dtype tf.int32, but got dtype tf.float32.

But seems like the problem is tensorflow constant function do not convert it properly to int32,

Consider changing line 1047 in sionna/mimo/detection.py to

 self._num_paths = tf.cast(tf.stack(num_paths, 0), tf.int32)

or line 1046 to

 num_paths.append(int(np.minimum(self._k, self._num_symbols**l)))

My tensorflow version is 2.10

[SebastianCa]

Hi, could you please provide a minimum example that reproduces the error? Which version of Sionna are you using and does upgrading to the latest version of TensorFlow solve the problem?

[jhoydis]

Closed due to inactivity.