from gym import make
import gym_workflow.envs
from collections import defaultdict, OrderedDict
import numpy as np
from agents.strategy.td import TD
import agents.utils.plotting as plt
if __name__ == '__main__':
env = make('Montage-v1')
episodes = 100
# Q Learning
Q, stats, records = TD.q_learning(
env,
episodes,
discount_factor=0.7,
epsilon=0.3,
training_episode=0,
log_file="exp1-cn-training-0-epi-{}-vm-10.csv".format(episodes),
)
sQ = OrderedDict(sorted(Q.items()), key=lambda i: keyorder.index(i[0]))
# plt.overhead_visualization(
# records['overhead'], xlabel='Cycle', ylabel="Overhead(s)",
# title="Overhead(s) across episodes"
# )
plt.v1_plot_action_value(
sQ, title="Experiment 1: Q-Value Overview - %s episodes" % episodes)
plt.v1_plot_episode_stats(stats)
if __name__ == '__main__':
env = make('Montage-v11')
episodes = 100
log_pre = "p2-exp1-epi-{}-train-0-maintain-all-terminal-20-M10-B20.csv".format(
episodes)
# log_pre = "exp-3-epi-{}-train-10-maintain-all".format(episodes)
# log_pre = "exp-3-epi-{}-train-0-maintain-smallest".format(episodes)
# log_pre = "exp-3-epi-{}-train-0-maintain-largest".format(episodes)
#
# Q Learning
Q, stats, records = TD.q_learning(
env,
episodes,
discount_factor=0.7,
epsilon=0.3,
training_episode=0,
log_file=log_pre,
)
sQ = OrderedDict(sorted(Q.items()), key=lambda i: keyorder.index(i[0]))
plt.plot_exp_2_action_value(sQ,
title="Q-Value Overview - %s episodes" %
episodes,
opt_lower=8,
opt_high=11,
xlim=11,
ylim=11)
plt.plot_episode_stats(stats, smoothing_window=10)
from collections import defaultdict, OrderedDict
import numpy as np
from agents.strategy.td import TD
import agents.utils.plotting as plt
if __name__ == '__main__':
# env = make('Montage-v8')
env = make('Montage-v12')
episodes = 500
# Q Learning
Q, stats, records = TD.q_learning(
env,
episodes,
discount_factor=0.7,
epsilon=0.3,
training_episode=0,
log_file="p3-exp1-training-epi-{}-vm-100-v12-B10-publication.csv".
format(episodes),
)
sQ = OrderedDict(sorted(Q.items()))
plt.plot_exp_3_action_value(
sQ, title="exp-4-v12-epi-{}-vm-100".format(episodes))
# plt.overall_records_visualization(records['benchmark'], xlabel="Cycle", ylabel="Benchmark Makespan(s)",
# title="Experiment 3: Benchmark Makespan over Cycle ({} Episodes)".format(
# episodes))
#
# plt.overall_records_visualization(
# records['makespan'], xlabel='Cycle', ylabel="Makespan(s)",
# title="Experiment 3: Makespan(s) across Cycle ({} Episodes)".format(episodes)
from gym import make
import gym_workflow.envs
from collections import defaultdict
import numpy as np
from agents.strategy.td import TD
import agents.utils.plotting as plt
if __name__ == '__main__':
env = make('Montage-v4')
episodes = 1000
Q, stats = TD.sarsa(env, episodes)
V = defaultdict(float)
for state, action_values in Q.items():
action_value = np.max(action_values)
V[state] = action_value
plt.plot_value_function(
V,
title="SARSA: Value Function representation - %s episodes" % episodes)
print(Q)
plt.plot_episode_stats(stats)