c7c06f56e918cabf565d4e4454daa344137d1f0f,contrib/rl/tictactoe.py,,main,#,151

Before Change

def main():
  env = TicTacToeEnvironment()
  policy = TicTacToePolicy()
  a3c = dc.rl.A3C(env, policy, entropy_weight=0, value_weight=0.25)
  a3c.optimizer = dc.models.tensorgraph.TFWrapper(
      tf.train.AdamOptimizer, learning_rate=0.01)
  a3c.fit(100000)
  env.reset()
  while not env._terminated:
    print(env.display())
    print(a3c.predict(env._state))

After Change

        print(value_weight)
        score = eval_tic_tac_toe(value_weight)
        scores[value_weight] = score
        with open("tictactoe_value_search.json", "w") as fout:
            fout.write(json.dumps(scores))
        value_weight += 0.05

Instances