Reusing replay_buffer with new model give KeyErrors on self.buffer[game_id]

[schwab]

Scenario

I'd like to reuse the replay_buffer from an earlier run, but on a slightly different model. Is there anyway to do this correctly? I've tried copying just the replay_buffer.pkl file to its own folder ./replays and just loading that folder with the new model, but this results in errors like ....

 (pid=547088, ip=192.168.1.175)
  File "python/ray/_raylet.pyx", line 463, in ray._raylet.execute_task
  File "python/ray/_raylet.pyx", line 415, in ray._raylet.execute_task.function_executor
  File "/home/mcstar_dev/project/muzero-general/replay_buffer.py", line 313, in reanalyse
    game_id, game_history, _ = ray.get(
ray.exceptions.RayTaskError(KeyError): ray::ReplayBuffer.sample_game() (pid=547087, ip=192.168.1.175)
  File "python/ray/_raylet.pyx", line 463, in ray._raylet.execute_task
  File "python/ray/_raylet.pyx", line 415, in ray._raylet.execute_task.function_executor
  File "/home/mcstar_dev/project/muzero-general/replay_buffer.py", line 155, in sample_game
    return game_id, self.buffer[game_id], game_prob
KeyError: -2

What's the correct way to accomplish re-using the old replay data for new runs or is this not supported? Is there anyway to force reanalyse to use only valid game_ids?

[ahainaut]

Closed this, since it's the same issue as #114