rl_agent/test_env.py

from env import Environment
from policy import Policy
from utils import myOptimizer

import pandas as pd
import numpy as np
import torch

if __name__ == "__main__":

    data = pd.read_csv('./data/EURUSD_Candlestick_1_M_BID_01.01.2021-04.02.2023.csv')
    # data['Local time'] = pd.to_datetime(data['Local time'])
    data = data.set_index('Local time')
    print(data.index.min(), data.index.max())

    date_split = '19.09.2022 17:55:00.000 GMT-0500'
    train = data[:date_split]
    test = data[date_split:]
    print(train.head(10))

    learning_rate = 0.01
    first_momentum = 0.0
    second_momentum = 0.0
    transaction_cost = 0.0001
    adaptation_rate = 0.01
    state_size = 8

    agent = Policy(input_channels=state_size).float()
    optimizer = myOptimizer(learning_rate, first_momentum, second_momentum, adaptation_rate, transaction_cost)

    

    history = []
    for i in range(1, state_size):
        c = train.iloc[i, :]['Close'] - train.iloc[i-1, :]['Close']
        history.append(c)

    env = Environment(train, history=history)
    observation = env.reset()

    for _ in range(9, 12):
        print(type(observation))
        observation = torch.as_tensor(observation)
        action = agent(observation)
        observation, reward, _ = env.step(action.data.numpy())

    print(env.profits)