search

AminHP / gym-anytrading

The most simple, flexible, and comprehensive OpenAI Gym trading environment (Approved by OpenAI Gym)
MIT License
2.13k stars 467 forks source link

On training how to set the frame_bound? #84

Closed talvasconcelos closed 1 year ago

talvasconcelos commented 1 year ago

I've been trying the stocksEnv on BTC data, to train a PPO agent. How do i set the frame_bound for training? it's the all training data, like frame_bound=(window_size, len(train)) ? or i set a smaller "window" that the bot learns on randomly, frame_bound=(window_size, 200)?

AminHP commented 1 year ago

Both methods or even a combination of them can be appropriate. Depending on the data and problem, each can have its own advantages. I think they are both worth investigating.

talvasconcelos commented 1 year ago

Also, when taking the saved model and trying to use it on a python script, what's the minimum requeirements to use the model to make predictions? Do i need gym and gym_anytrading in order to create an env? when loading the model, do always need an env?