robromijnders / rnn_basketball Goto Github PK

Hello again,

In line 56 of model.py, there is the line:
outputs, _ = tf.nn.rnn(cell, inputs, initial_state=initial_state)

To the best of my knowledge, tf.nn.rnn is not a real function. What should it be instead?

Thanks!

-Are you feeding one complete trajectory as a sequence ?
-If I set MDN to True then we are predicting the trajectory from a starting point to the basket right ?
-After training the model, how can I test the model (the trajectory prediction) ? Assume that I have a predefined starting point and goal point.
-How to predict a new trajectory after training ?

Thank you in advance.

Was running through the model and found a compilation error.

Traceback (most recent call last):
  File "main.py", line 85, in <module>
    model = Model(config)
  File "/home/neil/projects/RNN_basketball/model.py", line 101, in __init__
    normalize_theta = tf.inv(tf.reduce_sum(self.theta, 1, keep_dims=True))
AttributeError: 'module' object has no attribute 'inv'

It looks like inv was deprecated.

The network has [[0]] trainable parameters

???

Hello,

tf.unpack in model.py is deprecated, in favor of tf.unstack, which leads to an error. This is easily fixable for the user, but I thought you should know. Thanks!