Open jamiecl opened 4 years ago
Never mind it seems like it's generating the debug models which I think you renamed to model_t-dqn_GOOG_50. Just curious why does it generate a model for every 10 episodes if we just need 1 conclusive model to run eval.py with?
Also a side question, how can we make the model more accurate? Is it better to keep on retrain it using the same data or increase the number of episodes or something else?
Hi, I'm just trying to understand how the model is created I understand that you're repo comes with multiple models for GOOG with different episodes but how were these created? It seems for me after running train.py in the exact way you did except with data for SPY no model is instantiated.