dirs = [
ashvin_base_dir +
's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-wider3/run1',
ashvin_base_dir +
's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-wider3-relabeling/run1',
]
f = plot.filter_by_flat_params({
'algo_kwargs.num_updates_per_env_step':
4,
"replay_kwargs.fraction_goals_are_env_goals":
0.5,
"replay_kwargs.fraction_goals_are_rollout_goals":
0.2
})
exps = plot.load_exps(dirs, suppress_output=True)
plot.comparison(
exps,
["Final puck_distance Mean", "Final hand_distance Mean"],
vary=[
"replay_kwargs.fraction_goals_are_env_goals",
"replay_kwargs.fraction_goals_are_rollout_goals"
],
default_vary={"replay_strategy": "future"},
smooth=plot.padded_ma_filter(10),
xlim=(0, 500000),
ylim=(0.14, 0.26),
figsize=(6, 5),
method_order=[2, 1, 0, 3],
)
plt.gca().xaxis.set_major_formatter(plt.FuncFormatter(format_func))
import matplotlib.pyplot as plt
from rlkit.visualization import plot_util as plot
configure_matplotlib(matplotlib)
dirs = [
ashvin_base_dir +
's3doodad/ashvin/vae/fixed3/sawyer-pusher/state-dense-wider2/run1',
]
f = plot.filter_by_flat_params({
'algo_kwargs.num_updates_per_env_step':
4,
'replay_kwargs.fraction_goals_are_env_goals':
0.0
})
oracle = plot.load_exps(dirs, f, suppress_output=True)
plot.tag_exps(oracle, "name", "oracle")
dirs = [
ashvin_base_dir +
's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-wider3/run1',
ashvin_base_dir +
's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-wider3-relabeling/run1',
]
f = plot.filter_by_flat_params({
'algo_kwargs.num_updates_per_env_step':
4,
"replay_kwargs.fraction_goals_are_env_goals":
0.5,
"replay_kwargs.fraction_goals_are_rollout_goals":
0.2
import matplotlib
from visualization.grill.config import (
output_dir,
ashvin_base_dir,
format_func,
configure_matplotlib,
)
import matplotlib.pyplot as plt
from rlkit.visualization import plot_util as plot
configure_matplotlib(matplotlib)
f = plot.filter_by_flat_params(
{'replay_kwargs.fraction_goals_are_env_goals': 0.5})
exps = plot.load_exps(
[ashvin_base_dir + "s3doodad/share/reward-reaching-sweep"],
f,
suppress_output=True)
plot.comparison(
exps,
"Final distance Mean",
vary=["reward_params.type"],
# smooth=plot.padded_ma_filter(10),
ylim=(0.0, 0.2),
xlim=(0, 10000),
# method_order=[1, 0, 2]),
figsize=(6, 4),
)
plt.gca().xaxis.set_major_formatter(plt.FuncFormatter(format_func))
plt.xlabel("Timesteps")
plt.ylabel("Final Distance to Goal")
configure_matplotlib,
)
import matplotlib.pyplot as plt
from rlkit.visualization import plot_util as plot
configure_matplotlib(matplotlib)
f = plot.filter_by_flat_params({
'algo_kwargs.num_updates_per_env_step':
4,
'replay_kwargs.fraction_goals_are_env_goals':
0.5
})
oracle = plot.load_exps(
[ashvin_base_dir + "s3doodad/share/reacher/reacher-baseline-oracle"],
suppress_output=True)
plot.tag_exps(oracle, "name", "oracle")
ours = plot.load_exps(
[ashvin_base_dir + "s3doodad/share/reacher/reacher-main-results-ours"],
suppress_output=True)
plot.tag_exps(ours, "name", "ours")
f = plot.filter_by_flat_params({
'replay_kwargs.fraction_goals_are_env_goals': 0.0,
'reward_params.type': 'latent_distance'
})
her = plot.load_exps(
[ashvin_base_dir + "s3doodad/share/reward-reaching-sweep"],
f,
import matplotlib
from visualization.grill.config import (
output_dir,
ashvin_base_dir,
format_func,
our_method_name,
configure_matplotlib,
)
import matplotlib.pyplot as plt
from rlkit.visualization import plot_util as plot
configure_matplotlib(matplotlib)
# f = plot.filter_by_flat_params({'algorithm': 'Ours'})
ours = plot.load_exps([ashvin_base_dir + "s3doodad/share/real-reacher/ours"],
suppress_output=True)
plot.tag_exps(ours, "name", "ours")
# f = plot.filter_by_flat_params({'algorithm': 'Sparse-HER', 'reward_params.epsilon': 0.1})
her = plot.load_exps([ashvin_base_dir + "s3doodad/share/real-reacher/her"],
suppress_output=True)
plot.tag_exps(her, "name", "her")
# f = plot.filter_by_flat_params({'algorithm': 'TDM'})
oracle = plot.load_exps(
[ashvin_base_dir + "s3doodad/share/real-reacher/oracle"],
suppress_output=True)
plot.tag_exps(oracle, "name", "oracle")
plot.comparison(
ours + her + oracle,
"Test Final End Effector Distance from Target Mean",
vary=["name"],
import matplotlib
from visualization.grill.config import (
output_dir,
ashvin_base_dir,
format_func,
configure_matplotlib,
)
import matplotlib.pyplot as plt
from rlkit.visualization import plot_util as plot
configure_matplotlib(matplotlib)
exps = plot.load_exps([
ashvin_base_dir +
"s3doodad/share/reacher/reacher-abalation-resample-strategy"
],
suppress_output=True)
# plot.tag_exps(exps, "name", "oracle")
plot.comparison(
exps,
"Final distance Mean",
vary=[
"replay_kwargs.fraction_goals_are_env_goals",
"replay_kwargs.fraction_goals_are_rollout_goals"
],
# smooth=plot.padded_ma_filter(10),
ylim=(0.04, 0.2),
xlim=(0, 10000),
method_order=[2, 1, 0, 3],
figsize=(6, 4))
plt.gca().xaxis.set_major_formatter(plt.FuncFormatter(format_func))
import matplotlib
from visualization.grill.config import (
output_dir,
ashvin_base_dir,
format_func,
our_method_name,
configure_matplotlib,
)
import matplotlib.pyplot as plt
from rlkit.visualization import plot_util as plot
configure_matplotlib(matplotlib)
dirs = [
ashvin_base_dir + 's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-multi-logprob/run1', ]
logprob = plot.load_exps(dirs, suppress_output=True)
plot.tag_exps(logprob, "name", "logprob")
f = plot.filter_by_flat_params(
{'replay_kwargs.fraction_goals_are_env_goals': 0.5})
exps = plot.load_exps([
ashvin_base_dir + 's3doodad/share/steven/pushing-multipushing/multipusher-reward-variants'],
f, suppress_output=True)
plot.tag_exps(exps, "name", "pixel")
dirs = [
ashvin_base_dir + 's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-multi3/run1',
]
f = plot.filter_by_flat_params({
'algo_kwargs.num_updates_per_env_step': 4,
"replay_kwargs.fraction_goals_are_env_goals": 0.5
our_method_name,
configure_matplotlib,
)
import matplotlib.pyplot as plt
from rlkit.visualization import plot_util as plot
configure_matplotlib(matplotlib)
dirs = [
ashvin_base_dir + 's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-multi3/run1',
]
f = plot.filter_by_flat_params({
'algo_kwargs.num_updates_per_env_step': 4,
"replay_kwargs.fraction_goals_are_env_goals": 0.5
})
ours = plot.load_exps(dirs, f, suppress_output=True)
plot.tag_exps(ours, "name", "ours")
dirs = [
ashvin_base_dir + 's3doodad/ashvin/vae/fixed3/sawyer-pusher/state-dense-multi1/run1',
]
f = plot.filter_by_flat_params({
'replay_kwargs.fraction_goals_are_env_goals': 0.5,
'algo_kwargs.reward_scale': 1e-4
})
oracle = plot.load_exps(dirs, f, suppress_output=True)
plot.tag_exps(oracle, "name", "oracle")
# dsae = plot.load_exps(
# [ashvin_base_dir + 's3doodad/share/steven/pushing-multipushing/multipusher-reward-variants-spatial'],
# plot.filter_by_flat_params({'training_mode': 'test'}),
configure_matplotlib(matplotlib)
dirs = [
ashvin_base_dir +
's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-multi3/run1',
]
f = plot.filter_by_flat_params({
'algo_kwargs.num_updates_per_env_step':
4,
'rdim':
16,
'replay_kwargs.fraction_goals_are_rollout_goals':
0.2
})
her = plot.load_exps(dirs, f, suppress_output=True)
dirs = [
ashvin_base_dir +
's3doodad/ashvin/vae/fixed3/sawyer-pusher/vae-dense-multi3/run1',
]
f = plot.filter_by_flat_params({
'algo_kwargs.num_updates_per_env_step':
4,
'rdim':
16,
'replay_kwargs.fraction_goals_are_rollout_goals':
1.0,
'replay_kwargs.fraction_goals_are_env_goals':
0.0
})