def play(folder, nb_eps, seed, save=False):
"""
Replaying script.
Parameters
----------
folder: str
path to result folder.
nb_eps: int
Number of episodes to be replayed.
seed: int
save: bool
Whether to save figures.
"""
algorithm, cost_function, env, params = setup_for_replay(folder, seed)
goal = np.array([45000, 20])
for i_ep in range(nb_eps):
res, costs = algorithm.evaluate(n=1, best=False, goal=goal)#np.array([0.4]))#p.array([0.5, 1, 1]))
# print('----------------')
# for k in res.keys():
# print(k + ': {:.2f}'.format(res[k]))
stats = env.unwrapped.get_data()
labs = []
for l in stats['model_states_labels']:
if l == 'I':
labs.append(l)
else:
labs.append(l + r' $\mathbf{(\times 10^4)}$')
# plot model states
plot_stats(t=stats['history']['env_timesteps'],
states=np.array(stats['history']['model_states']).transpose() / np.array([1e4,1e4,1,1e4,1e4,1e4]).reshape(-1, 1),
labels=labs,#stats['model_states_labels'],
lockdown=np.array(stats['history']['lockdown']),
icu_capacity=stats['icu_capacity'],
time_jump=stats['time_jump'])
if save:
plt.savefig(folder + 'plots/model_states_ep_{}_{}.pdf'.format(i_ep, goal))
plt.close('all')
plot_stats(t=stats['history']['env_timesteps'][1:],
states=stats['stats_run']['to_plot'],
labels=stats['stats_run']['labels'],
legends=stats['stats_run']['legends'],
title=stats['title'],
lockdown=np.array(stats['history']['lockdown']),
time_jump=stats['time_jump'],
show=False if save else i_ep == (nb_eps - 1)
)
if save:
plt.savefig(folder + 'plots/rl_states_ep_{}_{}.pdf'.format(i_ep, goal))
plt.close('all')
simulation_horizon=simulation_horizon)) #params 이거 맞나?
env.reset()
actions = np.random.choice(
[0, 1, 2, 3, 4, 5],
size=53) # need to change to have 6 actions 0, 1, 2, 3, 4, 5
t = 0
r = 0
done = False
while not done:
out = env.step(actions[t])
t += 1
r += out[1]
done = out[2]
stats = env.unwrapped.get_data()
# plot model states
plot_stats(t=stats['history']['env_timesteps'],
states=np.array(stats['history']['model_states']).transpose(),
labels=stats['model_states_labels'],
distancing=np.array(stats['history']['distancing']),
time_jump=stats['time_jump'])
plot_stats(t=stats['history']['env_timesteps'][1:],
states=stats['stats_run']['to_plot'],
labels=stats['stats_run']['labels'],
legends=stats['stats_run']['legends'],
title=stats['title'],
distancing=np.array(stats['history']['distancing']),
time_jump=stats['time_jump'],
show=True)
print(r)
ax1 = None
ax2 = None
fig2 = None
# STEP 6: Plot results
for i in range(10):
stats = all_stats[i]
# plot model states
labs = [l + r' $(\times 10^3)$' for l in stats['model_states_labels']]
labs[2] = 'I'
ax1, fig1 = plot_stats(t=stats['history']['env_timesteps'],
states=np.array(stats['history']['model_states']).transpose(),
labels= stats['model_states_labels'],
lockdown=np.array(stats['history']['lockdown']),
time_jump=stats['time_jump'],
icu_capacity=stats['icu_capacity'],
axs=ax1)
plt.savefig('/home/flowers/Desktop/distrib.pdf')
ax2, fig2 = plot_stats(t=stats['history']['env_timesteps'][1:],
states=stats['stats_run']['to_plot'],
labels=stats['stats_run']['labels'],
title=stats['title'],
lockdown=np.array(stats['history']['lockdown']),
time_jump=stats['time_jump'],
axs=ax2,
fig=fig2
)
plt.show()
print(r)
ax1 = None
ax2 = None
fig2 = None
# STEP 6: Plot results
for i in range(10):
stats = all_stats[i]
# plot model states
labs = [l + r' $(\times 10^3)$' for l in stats['model_states_labels']]
labs[2] = 'I'
ax1, fig1 = plot_stats(
t=stats['history']['env_timesteps'],
states=np.array(stats['history']['model_states']).transpose(),
labels=stats['model_states_labels'],
distancing=np.array(stats['history']['distancing']),
time_jump=stats['time_jump'],
axs=ax1)
plt.savefig('/home/flowers/Desktop/distrib.pdf')
ax2, fig2 = plot_stats(t=stats['history']['env_timesteps'][1:],
states=stats['stats_run']['to_plot'],
labels=stats['stats_run']['labels'],
title=stats['title'],
distancing=np.array(
stats['history']['distancing']),
time_jump=stats['time_jump'],
axs=ax2,
fig=fig2)
plt.show()
model=model,
simulation_horizon=simulation_horizon)
env.reset()
actions = np.random.choice([0, 1], size=53)
t = 0
r = 0
done = False
while not done:
out = env.step(actions[t])
t += 1
r += out[1]
done = out[2]
stats = env.unwrapped.get_data()
# plot model states
plot_stats(t=stats['history']['env_timesteps'],
states=np.array(stats['history']['model_states']).transpose(),
labels=stats['model_states_labels'],
lockdown=np.array(stats['history']['lockdown']),
icu_capacity=stats['icu_capacity'],
time_jump=stats['time_jump'])
plot_stats(t=stats['history']['env_timesteps'][1:],
states=stats['stats_run']['to_plot'],
labels=stats['stats_run']['labels'],
legends=stats['stats_run']['legends'],
title=stats['title'],
lockdown=np.array(stats['history']['lockdown']),
time_jump=stats['time_jump'],
show=True)