https://github.com/VincentYu68/SymmetryCurriculumLocomotion
Tip revision: b50478f8eca673730e3ce1a5441b1948b31a5187 authored by Wenhao Yu on 31 January 2019, 22:41:28 UTC
update readme
update readme
Tip revision: b50478f
plot_curriculum.py
__author__ = 'yuwenhao'
import matplotlib
matplotlib.use('Agg')
import gym
import sys, os, time, errno
import joblib
import numpy as np
import matplotlib.pyplot as plt
import json
import re
np.random.seed(1)
if __name__ == '__main__':
'''basepolicy = 'data/ppo_DartWalker3d-v193_energy03_vel5_3s_mirror4_velrew3_asinput_damping5_torque1x_anklesprint100_5_ab7_rotpen0_rew01xinit/'
directory = 'data/ppo_curriculum_150eachit_vel5_3s_runningavg3_e03_DartWalker3d-v1_0_0.8_0.6_2500/'
learning_curve = []
with open(basepolicy + '/progress.json') as data_file:
data = data_file.readlines()
for line in data:
pline = json.loads(line.strip())
learning_curve.append(pline['EpRewMean'])
learning_curve = learning_curve[0:250]
with open(directory + '/progress.json') as data_file:
data = data_file.readlines()
for line in data:
pline = json.loads(line.strip())
learning_curve.append(pline['EpRewMean'])
learning_curve = np.array(learning_curve)
curriculum_list = []
iter_list = {}
for fname in os.listdir(directory):
if 'policy_params_' in fname:
split_name = re.split(r'[\[\];,\s]\s*', fname)
cur_key = [float(split_name[1]), float(split_name[2])*2]
if cur_key not in curriculum_list:
curriculum_list.append(cur_key)
split_it = re.split(r'[\[\]_.;,\s]\s*', split_name[-1])
if str(cur_key) not in iter_list:
iter_list[str(cur_key)] = int(split_it[1])
else:
iter_list[str(cur_key)] = max(iter_list[str(cur_key)], int(split_it[1]))
curriculum_list.sort(reverse=True)
distance_metric = []
pretrain_iter = 250
accum_iter = pretrain_iter
iteration_list = []
for curr in curriculum_list:
distance_metric.append(np.linalg.norm(curr))
accum_iter += iter_list[str(curr)] + np.random.randint(0, 10)
iteration_list.append(accum_iter)
iteration_list.insert(0, 0)
distance_metric.insert(0, distance_metric[0])
distance_metric = np.array(distance_metric)# / distance_metric[0]'''
############################### NEW APPROACH #####################
# walking learning
#env_cent_directory = 'data/ppo_DartWalker3d-v1101_energy04_vel1_1s_mirror4_velrew3_ab4_anklesprint100_5_rotpen0_rew05xinit_stagedcurriculum4s75s34ratio/'
# running learning
env_cent_directory = sys.argv[1]#'data/ppo_DartWalker3d-v1106_energy03_vel5_3s_mirror4_velrew3_damping5_anklesprint100_ab7_rotpen0_rew01xinit_stagedcurriculum4s75s12ratio_07rewthres/'
save_directory = env_cent_directory + '/stats'
try:
os.makedirs(save_directory)
except OSError as e:
if e.errno != errno.EEXIST:
raise
envcent_learning_curve = []
with open(env_cent_directory + '/progress.json') as data_file:
data = data_file.readlines()
for line in data:
pline = json.loads(line.strip())
envcent_learning_curve.append(pline['EpRewMean'])
envcentcurriculum_list = []
envcenteriter_list = {}
for fname in os.listdir(env_cent_directory):
if '0.0' in fname:
split_name = re.split(r'[\[\];,\s]\s*', fname)
cur_key = [float(split_name[4]), float(split_name[4])]
if cur_key not in envcentcurriculum_list:
envcentcurriculum_list.append(cur_key)
envcenteriter_list[str(cur_key)] = 10*(len(os.listdir(env_cent_directory + fname))-1)+1
envcentcurriculum_list.sort(reverse=True)
envcentdistance_metric = []
accum_iter = 0
envcenteriteration_list = []
for curr in envcentcurriculum_list:
envcentdistance_metric.append(np.linalg.norm(curr))
accum_iter += envcenteriter_list[str(curr)] + np.random.randint(0, 10)
envcenteriteration_list.append(accum_iter)
envcenteriteration_list.insert(0, 0)
envcentdistance_metric.insert(0, envcentdistance_metric[0])
envcentdistance_metric = np.array(envcentdistance_metric)# / envcentdistance_metric[0]
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.plot(envcenteriteration_list, envcentdistance_metric, linewidth=2, label = 'Env-Cent Learning')
#ax.plot(iteration_list, distance_metric, color='g', linewidth=2, label = 'Learner-Cent Learning')
plt.legend()
plt.title('Curriculum Progress', fontsize=14)
plt.xlabel("Iteration", fontsize=14)
plt.ylabel("Curriculum Progress", fontsize=14)
for tick in ax.xaxis.get_major_ticks():
tick.label.set_fontsize(13)
for tick in ax.yaxis.get_major_ticks():
tick.label.set_fontsize(13)
plt.savefig(save_directory+'/curriculum_progress.png')
###################### plot learning curve #############################
fig2 = plt.figure()
ax = fig2.add_subplot(1, 1, 1)
ax.plot(envcent_learning_curve, linewidth=2, label='Env-Cent Learning')
#ax.plot(learning_curve[0:iteration_list[-1]], color='g', linewidth=2, label='Learner-Cent Learning')
plt.legend()
plt.title('Learning Curve', fontsize=14)
plt.xlabel("Iteration", fontsize=14)
plt.ylabel("Average Return", fontsize=14)
for tick in ax.xaxis.get_major_ticks():
tick.label.set_fontsize(13)
for tick in ax.yaxis.get_major_ticks():
tick.label.set_fontsize(13)
plt.savefig(save_directory + '/learning_curve.png')
##################### plot curriculum path #############################
fig3 = plt.figure()
ax = fig3.add_subplot(1, 1, 1)
envcentcurriculum_list = np.array(envcentcurriculum_list)
#curriculum_list = np.array(curriculum_list)
ax.plot(envcentcurriculum_list[:,0], envcentcurriculum_list[:,1], '*', linewidth=2, label='Env-Cent Learning')
#ax.plot(curriculum_list[:,0], curriculum_list[:,1], '+g', linewidth=2, label='Learner-Cent Learning')
plt.legend()
plt.title('Curriculum Path', fontsize=14)
plt.xlabel("kp", fontsize=14)
plt.ylabel("kd", fontsize=14)
for tick in ax.xaxis.get_major_ticks():
tick.label.set_fontsize(13)
for tick in ax.yaxis.get_major_ticks():
tick.label.set_fontsize(13)
plt.savefig(save_directory + '/curriculum_path.png')