def test_no_3_reshape_future_many_times_are_rewards_ok(self):
# TRZEBA USTAWIC:
# if state.actual_phase == 'orange':
# return 'orange'
# return 0
# w metodzie get_action - wtedy ma przechodzic test
Globals().reshape_future = 4
agents=generate_random_epochs(learntAgents=False,epochs=range(50),u=env_settings.u_all_2) # bierze nowych agentow i tu jest 'is'
rewards = [mem.reward for mem in agents[0].memories]
a=14
from timeit import default_timer as timer
from nn_trainer import train
from random_epochs_generator import generate_random_epochs
from services.globals import Globals
from services.runnerService import run_learnt_greedy
Globals().pred_plot_memory = []
results = []
timeToLearn = 60
startTime = timer()
generate_random_epochs(learntAgents=False, epochs=range(50))
train(learntAgents=False)
run_learnt_greedy()
while timer() - startTime < timeToLearn:
print('czas', timer() - startTime)
generate_random_epochs(learntAgents=True, epochs=range(50))
train(max_time_learn=30)
result = run_learnt_greedy()
results.append(result)
from timeit import default_timer as timer
from nn_trainer import train
from random_epochs_generator import generate_random_epochs
from runner_learnt import run_learnt_greedy
from services.drawerService import draw_rewards_mean, draw_rewards, draw_cars_out
from services.globals import Globals
results = []
timeToLearn = 500000
startTime = timer()
agents = generate_random_epochs(learntAgents=False,
epochs=range(Globals().vp.first_epochs_range)) # bierze nowych agentow i tu jest 'is'
train(learntAgents=False, max_time_learn=Globals().vp.max_time_learn)
result = run_learnt_greedy()
eps_decay = 0
max_iterations_without_progress = 10
iterations_without_progress = 0
while timer() - startTime < timeToLearn:
eps_decay += 0.04
Globals().epsilon = 1 - eps_decay
if Globals().epsilon < 0.2:
Globals().epsilon = 0.2
print(
f'Czas:{round(timer() - startTime, 0)} Epsilon:{round(Globals().epsilon, 2)} Średnia liczba wjeżdżających pojazdów:'
f':{round(Globals().u_value, 2)}')
generate_random_epochs(learntAgents=True, epochs=range(Globals().vp.epochs_range))
train(max_time_learn=Globals().vp.max_time_learn)
result = run_learnt_greedy()
maximum_possible_cars_out = result[4]
if result[2] > maximum_possible_cars_out * 0.93: # cars_out
Globals().u_value = Globals().u_value * 1.2
from nn_trainer import train
from random_epochs_generator import generate_random_epochs
from runner_learnt import run_learnt_greedy
import matplotlib.pyplot as plt
from services.globals import Globals
runs = [0, 1, 2]
for run in runs:
Globals().pred_plot_memory = []
Globals().run_no = run
results = []
timeToLearn = 600
startTime = timer()
generate_random_epochs(
learntAgents=False,
epochs=range(50)) # bierze nowych agentow i tu jest 'is'
train()
run_learnt_greedy()
while timer() - startTime < timeToLearn:
print('czas', timer() - startTime)
generate_random_epochs(learntAgents=True, epochs=range(50))
train(max_time_learn=40)
result = run_learnt_greedy()
results.append(result)
# print(times)
plt.plot([res[0] for res in results])
plt.title('Średnia wszystkich nagród - akcje wedle wyuczonej strategii')
plt.savefig('rewards_mean' + str(run) + '.png')
plt.close()
# plt.show() # rewards_mean
# agents=generate_random_epochs(learntAgents=False,
# epochs=range(Globals().vp().first_epochs_range)) # bierze nowych agentow i tu jest 'is'
# # draw_colored_batches(agents)
# train(learntAgents=False, max_time_learn=Globals().vp().max_time_learn)
# result=run_learnt_greedy()
lurns = 0
eps_decay = 0
while timer() - startTime < timeToLearn:
eps_decay += 0.04
Globals().epsilon = 1 - eps_decay
if Globals().epsilon < 0.2:
Globals().epsilon = 0.2
print('epsilon', Globals().epsilon)
print('czas', timer() - startTime)
print('U', Globals().u_value)
generate_random_epochs(learntAgents=True,
epochs=range(Globals().vp().epochs_range))
draw_batches('static_files/x_batch' + str(Globals().greedy_run_no))
train(max_time_learn=Globals().vp().max_time_learn)
result = run_learnt_greedy()
draw_batches_from_agent(result[3],
file_name='static_files/x_batch_greedy' +
str(Globals().greedy_run_no))
maximum_possible_cars_out = Globals().u_value * Globals().vp(
).max_time_greedy * 8
print('max possible', maximum_possible_cars_out)
if result[2] > maximum_possible_cars_out * 0.93: # cars_out
print('u przed', Globals().u_value)
Globals().u_value = Globals().u_value * 1.2
# print('u po',Globals().u_value)
results.append(result)
lurns += 1