Cannot feed value of shape (4, 198) for Tensor 'agent_BM_Ask/obs:0', which has shape '(?, 177)

RLlib
2

Hi @carlorop,

One thing that seems off to me is this:

<tf.Tensor 'agent_BM_Ask/timestep:0' shape=() dtype=int64>: 96}

That seems to indicate that there were at least 96 previous steps that succeeded.

Do you have some kind of conditional in your environment that determines how your observation is created?

Is it possible that there is some condition in the environment that forms the observation incorrectly?

Are you using a custom model or one of the RLlib built in models?

1 Like