1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128 | """Argument parser and logic for the ``plot_train`` sub-command."""
from __future__ import annotations
import argparse
import os
from pathlib import Path
import numpy as np
try:
from matplotlib import pyplot as plt
import seaborn # ty:ignore[unresolved-import]
_PLOT_DEPS_AVAILABLE = True
except ImportError:
_PLOT_DEPS_AVAILABLE = False
def _add_plot_train_args(parser: argparse.ArgumentParser) -> None:
parser.add_argument("-a", "--algo", help="Algorithm to include", type=str, required=True)
parser.add_argument("-e", "--env", help="Environment(s) to include", nargs="+", type=str, required=True)
parser.add_argument("-f", "--exp-folder", help="Folders to include", type=str, required=True)
parser.add_argument(
"--figsize", help="Figure size, width, height in inches.", nargs=2, type=int, default=[6.4, 4.8]
)
parser.add_argument("--fontsize", help="Font size", type=int, default=14)
parser.add_argument("-max", "--max-timesteps", help="Max number of timesteps to display", type=int)
parser.add_argument(
"-x",
"--x-axis",
help="X-axis",
choices=["steps", "episodes", "time"],
type=str,
default="steps",
)
parser.add_argument(
"-y",
"--y-axis",
help="Y-axis",
choices=["success", "reward", "length"],
type=str,
default="reward",
)
parser.add_argument("-w", "--episode-window", help="Rolling window size", type=int, default=100)
def run_plot_train(args: argparse.Namespace) -> None:
"""Execute the plot_train command with pre-parsed args."""
if not _PLOT_DEPS_AVAILABLE:
raise ImportError(
"matplotlib and seaborn are required for plotting. Install them with: pip install matplotlib seaborn"
)
from lit_sb3.core.results_plotter import X_EPISODES, X_TIMESTEPS, X_WALLTIME, ts2xy, window_func
from lit_sb3.wrappers.monitor import LoadMonitorResultsError, load_results
# Activate seaborn
seaborn.set()
algo = args.algo
envs = args.env
log_path = Path(args.exp_folder) / algo
x_axis = {
"steps": X_TIMESTEPS,
"episodes": X_EPISODES,
"time": X_WALLTIME,
}[args.x_axis]
x_label = {
"steps": "Timesteps",
"episodes": "Episodes",
"time": "Walltime (in hours)",
}[args.x_axis]
y_axis = {
"success": "is_success",
"reward": "r",
"length": "l",
}[args.y_axis]
y_label = {
"success": "Training Success Rate",
"reward": "Training Episodic Reward",
"length": "Training Episode Length",
}[args.y_axis]
dirs = []
for env in envs:
# Sort by last modification
entries = sorted(os.scandir(log_path), key=lambda entry: entry.stat().st_mtime)
dirs.extend(entry.path for entry in entries if env in entry.name and entry.is_dir())
plt.figure(y_label, figsize=args.figsize)
plt.title(y_label, fontsize=args.fontsize)
plt.xlabel(f"{x_label}", fontsize=args.fontsize)
plt.ylabel(y_label, fontsize=args.fontsize)
for folder in dirs:
try:
data_frame = load_results(folder)
except LoadMonitorResultsError:
continue
if args.max_timesteps is not None:
data_frame = data_frame[data_frame.l.cumsum() <= args.max_timesteps]
try:
y = np.array(data_frame[y_axis])
except KeyError:
print(f"No data available for {folder}")
continue
x, _ = ts2xy(data_frame, x_axis)
# Do not plot the smoothed curve at all if the timeseries is shorter than window size.
if x.shape[0] >= args.episode_window:
# Compute and plot rolling mean with window of size args.episode_window
x, y_mean = window_func(x, y, args.episode_window, np.mean)
plt.plot(x, y_mean, linewidth=2, label=folder.split("/")[-1])
plt.legend()
plt.tight_layout()
plt.show()
def plot_train() -> None:
"""Plot training reward or success rate (parse args then run)."""
parser = argparse.ArgumentParser("Gather results, plot training reward/success")
_add_plot_train_args(parser)
args = parser.parse_args()
run_plot_train(args)
|