I've modularized my code and I have now a separate script for training/testing a gym RL model then saving the model as .zip and another script for loading the .zip model into an env initialized with new data including unseen data. I have used the same parameters (window, episodes, frame bounds, train/test split, env type etc) in training/testing and loading but it keeps throwing an error that my observation spaces do not match:
ValueError(f"Observation spaces do not match: {observation_space} != {env.observation_space}")
Box(-inf, inf, (AA,BB), float32) != Box(-inf, inf, (CC,BB),float32)
My env observation space contains the correct parameters CC and BB but they should be CC and BB in the "observation_space". I'm trying to figure out where the AA value came from and why its not equal to CC. How can we call the "observation_space" from a saved gym .zip model? I need to find out where it comes from.....
Hi Amin,
I've modularized my code and I have now a separate script for training/testing a gym RL model then saving the model as .zip and another script for loading the .zip model into an env initialized with new data including unseen data. I have used the same parameters (window, episodes, frame bounds, train/test split, env type etc) in training/testing and loading but it keeps throwing an error that my observation spaces do not match:
My env observation space contains the correct parameters CC and BB but they should be CC and BB in the "observation_space". I'm trying to figure out where the AA value came from and why its not equal to CC. How can we call the "observation_space" from a saved gym .zip model? I need to find out where it comes from.....
Thanks!