def plot_contour(train_dir):
ant_train_points = Path(train_dir)
env_name = os.path.basename(ant_train_points)
generated_dirs_dir = f"generated_dirs/_temp_dir/"
if os.path.exists(generated_dirs_dir):
shutil.rmtree(generated_dirs_dir)
checkpoint_names = [checkpoint for checkpoint in os.listdir(ant_train_points) if checkpoint[-1] == '0']
checkpoint_names.sort()
vectors = []
for checkpoint in (checkpoint_names):
check_param_fname = ant_train_points / checkpoint / "parameters.th"
params = torch.load(check_param_fname,map_location=torch.device('cpu')).values()
dir1 = [v.detach().flatten() for v in params]
vector = torch.cat(dir1,axis=0).numpy()
vectors.append(vector)
decomposer = sklearn.decomposition.PCA(2)
coords = decomposer.fit_transform(vectors)
# center points at final coordinate, i.e. center of plot
coords -= coords[-1]
dir1, dir2 = decomposer.components_
dir1 = npvec_to_nplist(dir1, params)
dir2 = npvec_to_nplist(dir2, params)
scale1,scale2 = np.max(np.abs(coords),axis=0)*1.15
dir1 = [d*scale1 for d in dir1]
dir2 = [d*scale2 for d in dir2]
num_episodes = 25
train_info = json.load(open(ant_train_points/"info.json"))
base_checkpoint = str(ant_train_points / checkpoint_names[-1])
generate_plane_data(base_checkpoint, generated_dirs_dir, dir1, dir2, train_info, grid_size=15, num_episodes=num_episodes)
run_job_list(generated_dirs_dir+"jobs.sh")
job_results_to_csv(generated_dirs_dir)
plot_plane(generated_dirs_dir+"results.csv",
outname="contours/"+env_name,
type="contour",
dir1_name="component 1",
dir2_name="component 2",
dir1_scale=scale1,
dir2_scale=scale2,
points=coords,
)
def replot(csv_fname,
outname=None,
env_name=None,
key_name="episode_rewards",
title=None,
plot_type="mesh",
file_type="png",
logscale=False):
default_outname = "vis/" + "".join(
[c for c in csv_fname if re.match(r'\w', c)]) + key_name
outname = outname if outname is not None else default_outname
datafname = csv_fname
plot_plane(datafname,
outname,
key_name=args.key,
plot_type=plot_type,
file_type=file_type,
env_name=env_name,
logscale=False)
parser.add_argument('--dir1_name',
type=str,
help="Name of the x axis in the plot")
parser.add_argument('--dir2_name',
type=str,
help="Name of the y axis in the plot")
parser.add_argument(
'--show',
action='store_true',
help="Shows plot instead of saving it (does not work for vtp output)")
parser.add_argument(
'--logscale',
default="auto",
help=
"Plot in log scale rather than the default linear scale. One of 'auto', 'on', or 'off'"
)
args = parser.parse_args()
plot_plane(args.datafname,
args.outname,
env_name=args.env_name,
key_name=args.key,
plot_type=args.type,
dir1_name=args.dir1_name,
dir2_name=args.dir2_name,
dir1_scale=args.dir1_scale,
dir2_scale=args.dir2_scale,
show=args.show,
logscale=args.logscale)
from reward_surfaces.experiments import generate_plane_data
from reward_surfaces.plotting import plot_plane
from reward_surfaces.runners.run_jobs_multiproc import run_job_list
from reward_surfaces.utils.job_results_to_csv import job_results_to_csv
from reward_surfaces.utils.surface_utils import readz
from reward_surfaces.agents.make_agent import make_agent
import json
import os
import shutil
trained_checkpoint = "train_results/bullet/hopper/0040000/"
generated_dirs_dir = "generated_dirs/test_script/"
dir1_fname = "generated_dirs/hopper_eig_vecs/results/0040000/mineigvec.npz"
dir2_fname = "generated_dirs/hopper_eig_vecs/results/0040000/maxeigvec.npz"
dir1 = readz(dir1_fname)
dir2 = readz(dir2_fname)
train_info = json.load(open(trained_checkpoint+"../info.json"))
if os.path.exists(generated_dirs_dir):
shutil.rmtree(generated_dirs_dir)
print("removed")
generate_plane_data(trained_checkpoint, generated_dirs_dir, dir1, dir2, train_info, num_steps=1000)
run_job_list(generated_dirs_dir+"jobs.sh")
job_results_to_csv(generated_dirs_dir)
plot_plane(generated_dirs_dir+"results.csv")
job_results_to_csv(gif_data_folder / checkpoint)
base_mag = None
factor = 2
frame_idx = 0
for checkpoint in checkpoints:
csv_path = gif_data_folder / checkpoint / "results.csv"
row_dat = pandas.read_csv(csv_path)[args.key]
m = max(0.1, np.max(row_dat))
mini = min(0, np.min(row_dat))
print(m)
if base_mag is None:
base_mag = m
else:
while m > base_mag:
base_mag = base_mag * factor
fname = plot_plane(str(csv_path),
str(frames_dir + checkpoint),
env_name=args.env_name,
key_name=args.key,
plot_type=args.type,
vmin=mini,
vmax=base_mag,
show=False,
logscale="off")
if not fname:
continue
os.rename(fname, f"{frames_dir}{frame_idx:05}.png")
frame_idx += 1