def test_message_float_list():
val = [1.0, 3.0, 9.0]
msg_out = OutgoingMessage()
msg_out.write_float32_list(val)
msg_in = IncomingMessage(msg_out.buffer)
read_val = msg_in.read_float32_list()
# These won't be exactly equal in general, since python floats are 64-bit.
assert val == read_val
# Test reading with defaults
assert [] == msg_in.read_float32_list()
assert val == msg_in.read_float32_list(default_value=val)
def on_message_received(self, msg: IncomingMessage) -> None:
agent_id = msg.read_int32()
step = msg.read_int32()
new_rewards = msg.read_float32_list()
if agent_id in self._rewards:
last_step, last_rewards = self._rewards[agent_id]
assert last_step == -1 or last_step == step
for i in range(len(new_rewards)):
new_rewards[i] += last_rewards
self._rewards[agent_id] = (step, new_rewards)
