init.py

from src.WorldModel import Enve 
from src.ComputeEngine import DQN_Actor, DoubleDQN_Actor, Linear_Programming, EMA_Online

import torch
import numpy as np
import matplotlib.pyplot as plt

def train_QNetwork(n_games):	
	game_scores, game_losses = [], []

	for i in range(n_games):
		observation, reward, done = env.reset()
		score, counter = 0, 0

		while not env.done:
			action = agent.choose_action(observation)
			observation_, reward, done = env.step(action)
			score += reward

			agent.store_transition(observation, action, reward, observation_, done)
			agent.learn()
			observation = observation_
			counter+=1

		game_scores.append(score)
		avg_score = np.mean(game_scores[-100:], dtype=float)

		if i>10:
			if (i+1) % 10 == 0:
				agent.writer.add_scalar("Avg. Score", avg_score, i+1)
				print("Game: {} AvgScore: {:.3f}".format(i+1, avg_score))
			agent.writer.add_scalar("Score", score, i)

		if (i+1) % 4000 == 0:
			agent.save_ckpt(i+1, "{:.3f}".format(avg_score))
	agent.writer.close()

def eval_QNetwork(model_path, startIndex=100, endIndex=130, soc=0.6):

	observation, reward, done = env.test(startIndex, endIndex, soc)
	agent.load_ckpt(model_path)

	out = []
	out.append(observation)
	score=0
	while not done:
		action = agent.choose_action(observation, test=True)
		print(f"Observation: {observation} | Action: {action} ")
		observation_, reward, done = env.step(action)
		score += reward
		out.append(observation_)
		observation = observation_

	out = np.asarray(out)
	return out, score

if __name__ == "__main__":
	data_file = "Data/PriceData.csv"
	

	env 	= Enve(DataFile_path=data_file, max_charge=0.8, min_charge=0.2, rate=0.1, battery_cap=1500, \
								strategy_no=3, ema_coeff=-0.05, ema_len=6, max_episode_len=200, min_episode_len=100)
	
	agent = DoubleDQN_Actor(gamma=0.99, epsilon=1, lr=0.001, input_dims=4, batch_size=32, num_actions=3)
	train_QNetwork(10_000)
	
	agent = DQN_Actor(gamma=0.99, epsilon=1, lr=0.001, input_dims=4, batch_size=32, num_actions=3)
	train_QNetwork(10_000)
	
	print("...done...")


	# ema_ = EMA_Online( data_file, maxCharge=0.8, minCharge=0.2, rate=0.1 , batteryCap =100)
	# ema_.run(soc=0.6, startIndex=100, endIndex=150, window=6)
	# python3 -m tensorboard.main --logdir=~/my/training/dir