lil-lab / blocks Goto Github PK

Traceback (most recent call last):
  File "./experiments/test_contextualbandit.py", line 12, in <module>
    agent = Agent(CONTEXTUALBANDIT, config, constants_hyperparam)
  File "/home/anirban/Documents/Repos/blocks/BlockWorldRoboticAgent/agent.py", line 83, in __init__
    self.model = PolicyNetwork(image_dim, self.num_actions, constants)
  File "/home/anirban/Documents/Repos/blocks/BlockWorldRoboticAgent/model/policy_network.py", line 21, in __init__
    self.text_embedder = embed_token_seq.EmbedTokenSeq(self.n_text)
  File "/home/anirban/Documents/Repos/blocks/BlockWorldRoboticAgent/model/embed_token_seq.py", line 58, in __init__
    self._initial_state = cell.zero_state(self.batch_size, tf.float32)
  File "/home/anirban/.local/lib/python2.7/site-packages/tensorflow_core/python/ops/rnn_cell_impl.py", line 1265, in zero_state
    return tuple(cell.zero_state(batch_size, dtype) for cell in self._cells)
  File "/home/anirban/.local/lib/python2.7/site-packages/tensorflow_core/python/ops/rnn_cell_impl.py", line 1265, in <genexpr>
    return tuple(cell.zero_state(batch_size, dtype) for cell in self._cells)
  File "/home/anirban/.local/lib/python2.7/site-packages/tensorflow_core/python/ops/rnn_cell_impl.py", line 337, in zero_state
    output = _zero_state_tensors(state_size, batch_size, dtype)
  File "/home/anirban/.local/lib/python2.7/site-packages/tensorflow_core/python/ops/rnn_cell_impl.py", line 179, in _zero_state_tensors
    return nest.map_structure(get_state_shape, state_size)
  File "/home/anirban/.local/lib/python2.7/site-packages/tensorflow_core/python/util/nest.py", line 536, in map_structure
    structure[0], [func(*x) for x in entries],
  File "/home/anirban/.local/lib/python2.7/site-packages/tensorflow_core/python/ops/rnn_cell_impl.py", line 172, in get_state_shape
    c = _concat(batch_size, s)
  File "/home/anirban/.local/lib/python2.7/site-packages/tensorflow_core/python/ops/rnn_cell_impl.py", line 134, in _concat
    "but saw tensor: %s" % p)
ValueError: prefix tensor must be either a scalar or vector, but saw tensor: Tensor("Placeholder:0", dtype=int32)