Alexey Maslov

@dvdhfnr how are things with implementing your tf agents trained NN on Android? I have this error: "RuntimeError: Encountered unresolved custom op: BroadcastArgs.Node number 0 (BroadcastArgs) failed to prepare." Here...

@dvdhfnr you are right, the problem is this ops: ``` Exception: :0: error: loc(fused["Deterministic_1/sample/BroadcastArgs@__inference_action_11129549", "StatefulPartitionedCall/StatefulPartitionedCall/StatefulPartitionedCall/StatefulPartitionedCall/Deterministic_1/sample/BroadcastArgs"]): 'tf.BroadcastArgs' op is neither a custom op nor a flex op :0: error: loc(fused["ActorDistributionNetwork/TanhNormalProjectionNetwork/MultivariateNormalDiag/shapes_from_loc_and_scale/prefer_static_broadcast_shape/BroadcastArgs@__inference_action_11129549", "StatefulPartitionedCall/StatefulPartitionedCall/StatefulPartitionedCall/StatefulPartitionedCall/ActorDistributionNetwork/TanhNormalProjectionNetwork/MultivariateNormalDiag/shapes_from_loc_and_scale/prefer_static_broadcast_shape/BroadcastArgs"]):...