def load():
if progress_filename == "progress.csv":
return core.load_exps_data(dirnames)
elif progress_filename == "tensorboard_log.npy":
return load_exps_data_numpy(dirnames, progress_filename=progress_filename)
else:
return core.load_exps_data(dirnames, progress_filename=progress_filename)
def load():
if progress_filename == "progress.csv":
return core.load_exps_data(dirnames)
elif progress_filename == "tensorboard_log.npy":
return load_exps_data_numpy(dirnames, NumpyLogReader(), progress_filename=progress_filename, )
elif custom_log_reader:
return load_exps_data_numpy(dirnames, custom_log_reader, progress_filename=progress_filename, )
else:
return core.load_exps_data(dirnames, progress_filename=progress_filename)
def load_exps(dirnames, filter_fn=true_fn, suppress_output=False):
if suppress_output:
with suppress_stdout():
exps = core.load_exps_data(dirnames)
else:
exps = core.load_exps_data(dirnames)
good_exps = []
for e in exps:
if filter_fn(e):
good_exps.append(e)
return good_exps
def reload_data(data_paths):
"""
Iterate through the data folder and organize each experiment into a list, with their progress data, hyper-parameters
and also analyze all the curves and give the distinct hyper-parameters.
:param data_path: Path of the folder storing all the data
:return [exps_data, plottable_keys, distinct_params]
exps_data: A list of the progress data for each curve. Each curve is an AttrDict with the key
'progress': A dictionary of plottable keys. The val of each key is an ndarray representing the
values of the key during training, or one column in the progress.txt file.
'params'/'flat_params': A dictionary of all hyperparameters recorded in 'variants.json' file.
plottable_keys: A list of strings representing all the keys that can be plotted.
distinct_params: A list of hyper-parameters which have different values among all the curves. This can be used
to split the graph into multiple figures. Each element is a tuple (param, list_of_values_to_take).
"""
exps_data = copy.copy(
core.load_exps_data(data_paths,
disable_variant=False,
ignore_missing_keys=True))
plottable_keys = copy.copy(
sorted(
list(set(flatten(list(exp.progress.keys())
for exp in exps_data)))))
distinct_params = copy.copy(sorted(
core.extract_distinct_params(exps_data)))
return exps_data, plottable_keys, distinct_params
def reload_data():
global exps_data
global plottable_keys
global distinct_params
exps_data = core.load_exps_data(args.data_paths, args.disable_variant)
plottable_keys = sorted(list(
set(flatten(list(exp.progress.keys()) for exp in exps_data))))
distinct_params = sorted(core.extract_distinct_params(exps_data))
def reload_data():
global exps_data
global plottable_keys
global distinct_params
exps_data = core.load_exps_data(args.data_paths,args.disable_variant)
plottable_keys = sorted(list(
set(flatten(list(exp.progress.keys()) for exp in exps_data))))
distinct_params = sorted(core.extract_distinct_params(exps_data))
def reload_data():
global exps_data
global plottable_keys
global distinct_params
exps_data = core.load_exps_data(args.data_path)
plottable_keys = list(
set(flatten(exp.progress.keys() for exp in exps_data)))
distinct_params = core.extract_distinct_params(exps_data)
def reload_data():
global exps_data
global plottable_keys
global distinct_params
exps_data = core.load_exps_data(args.data_paths,args.disable_variant)
plottable_keys = list(
set(flatten(list(exp.progress.keys()) for exp in exps_data)))
plottable_keys = sorted([k for k in plottable_keys if k is not None])
# distinct_params = sorted(core.extract_distinct_params(exps_data))
distinct_params = [] # todo: to get normal plots at all
def reload_data():
global exps_data
global plottable_keys
global distinct_params
args.disable_variant = False
exps_data = core.load_exps_data(args.data_paths, args.disable_variant)
plottable_keys = list(
set(flatten(list(exp.progress.keys()) for exp in exps_data)))
plottable_keys = sorted([k for k in plottable_keys if k is not None])
distinct_params = sorted(core.extract_distinct_params(exps_data))
print("\n\n distinct_params:{} \n\n".format(distinct_params))
plt.rc('axes', labelsize=MEDIUM_SIZE) # fontsize of the x and y labels
plt.rc('xtick', labelsize=SMALL_SIZE) # fontsize of the tick labels
plt.rc('ytick', labelsize=SMALL_SIZE) # fontsize of the tick labels
plt.rc('legend', fontsize=MEDIUM_SIZE) # legend fontsize
plt.rc('figure', titlesize=BIGGER_SIZE) # fontsize of the figure title
prop_cycle = plt.rcParams['axes.prop_cycle']
colors = prop_cycle.by_key()['color']
COLORS = dict(ours=colors.pop(0))
LEGEND_ORDER = {'ppo': 0, 'trpo': 1, 'vpg': 2}
########## Add data path here #############
data_path = "data/s3/ppo-maml-hyperparam-final-2-lmao"
###########################################
exps_data = core.load_exps_data([data_path], False)
def sorting_legend(label):
return LEGEND_ORDER[label]
def get_color(label):
if label not in COLORS.keys():
COLORS[label] = colors.pop(0)
return COLORS[label]
def total_under_curve(exp_data):
mean_scores = []
for exp in exp_data:
def load():
if progress_filename == "progress.csv":
return core.load_exps_data(dirnames)
else:
return core.load_exps_data(dirnames, progress_filename=progress_filename)
from rllab.viskit import frontend
from rllab.viskit import core
import matplotlib.pyplot as plt
from plots.plot_utils import *
plt.style.use('ggplot')
#plt.rc('font', family='Times New Roman')
import matplotlib
# matplotlib.use('TkAgg')
#matplotlib.font_manager._rebuild()
data_path_fast_lr = None
data_path_num_models = None
data_path_maml_iter = None
exps_data_fast_lr = core.load_exps_data([data_path_fast_lr], False)
exps_data_num_models = core.load_exps_data([data_path_num_models], False)
exps_data_maml_iter = core.load_exps_data([data_path_maml_iter], False)
SMALL_SIZE = 30
MEDIUM_SIZE = 32
BIGGER_SIZE = 36
LINEWIDTH = 4
plt.rc('font', size=SMALL_SIZE) # controls default text sizes
plt.rc('axes', titlesize=BIGGER_SIZE) # fontsize of the axes title
plt.rc('axes', labelsize=MEDIUM_SIZE) # fontsize of the x and y labels
plt.rc('xtick', labelsize=SMALL_SIZE) # fontsize of the tick labels
plt.rc('ytick', labelsize=SMALL_SIZE) # fontsize of the tick labels
plt.rc('legend', fontsize=SMALL_SIZE) # legend fontsize
plt.rc('figure', titlesize=BIGGER_SIZE) # fontsize of the figure title
