Issue with the StockTradingEnv for 1m ticker data

[lokeshsoni]

I've been trying to train a model with 1m of tick data for the DOW_30 Index. Everything works perfectly fine for daily data, but when I do the same steps for 1m, I get errors. I've resolved the errors in the feature engineering by making modifications locally, as it looks like there was no support for 1m data.

raw data - https://www.dropbox.com/s/sdhzjl7jv5oaada/raw_dow30_2022-09-11_2022-09-18_1m.csv?dl=0 feature engineered data (without turbulence index) - https://www.dropbox.com/s/ohtdcsh01e88j3y/fe_dow30_2022-09-11_2022-09-18_1m.csv?dl=0

from finrl.config import INDICATORS
from finrl.meta.env_stock_trading.env_stocktrading import StockTradingEnv

stock_dimension = len(train.tic.unique())
state_space = 1 + 2*stock_dimension + len(INDICATORS)*stock_dimension
buy_cost_list = sell_cost_list = [0.001] * stock_dimension
num_stock_shares = [0] * stock_dimension

env_kwargs = {
    "hmax": 100,
    "initial_amount": 1000000,
    "num_stock_shares": num_stock_shares,
    "buy_cost_pct": buy_cost_list,
    "sell_cost_pct": sell_cost_list,
    "state_space": state_space,
    "stock_dim": stock_dimension,
    "tech_indicator_list": INDICATORS,
    "action_space": stock_dimension,
    "reward_scaling": 1e-4
}

e_train_gym = StockTradingEnv(df = train, **env_kwargs)

Error - AttributeError: 'numpy.float64' object has no attribute 'values'

On digging deeper, it looks like the error is coming while initializing the state in the StockTradingEnv class under _initiate_state function. It looks like it's expecting an array self.data.close.values.tolist() but as per the initial declaration self.data = self.df.loc[self.day, :], a scalar value is coming.

Please let me know if I'm doing anything wrong. Also, help me with any existing implementation on 1m data.

Thank you Lokesh Soni

[sebnapi]

Hey Lokesh, did you solve it yourself? The code is still the same.