File size: 3,448 Bytes
047bd6f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import json
import re
from pathlib import Path

import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt

RUNS = Path("runs")
OUT = Path("results")
OUT.mkdir(exist_ok=True)


def load_runs():
    runs = []
    for d in sorted(RUNS.iterdir()):
        hist = d / "history.json"
        if not hist.exists():
            continue
        m = re.fullmatch(r"depth(\d+)(_res)?(_nobn)?", d.name)
        depth = int(m.group(1))
        residual = bool(m.group(2))
        bn = not m.group(3)
        data = json.loads(hist.read_text())
        runs.append({
            "name": d.name, "depth": depth, "residual": residual, "bn": bn,
            "epochs": [h["epoch"] for h in data],
            "train_loss": [h["train_loss"] for h in data],
            "train_acc": [h["train_acc"] for h in data],
            "val_acc": [h["val_acc"] for h in data],
        })
    return runs


def style(r):
    color = "tab:blue" if not r["residual"] else "tab:red"
    ls = "-" if r["bn"] else "--"
    return color, ls


def label(r):
    return f"depth{r['depth']}" + ("-res" if r["residual"] else "")


def plot_curves(runs, bn, metric, ylabel, fname, logy=False):
    fig, ax = plt.subplots(figsize=(8, 5))
    for r in sorted(runs, key=lambda r: (r["depth"], r["residual"])):
        if r["bn"] != bn:
            continue
        c, ls = style(r)
        ax.plot(r["epochs"], r[metric], ls, color=c, label=label(r), lw=1.5)
    ax.set_xlabel("epoch")
    ax.set_ylabel(ylabel)
    ax.set_title(f"{ylabel} vs epoch ({'with' if bn else 'without'} BatchNorm)")
    ax.legend(fontsize=8)
    ax.grid(alpha=0.3)
    if logy:
        ax.set_yscale("log")
    fig.tight_layout()
    fig.savefig(OUT / fname, dpi=150)
    plt.close(fig)


def plot_summary(runs, fname):
    fig, axes = plt.subplots(1, 2, figsize=(11, 4))
    for ax, bn, title in [(axes[0], True, "with BN"), (axes[1], False, "without BN")]:
        subset = [r for r in runs if r["bn"] == bn]
        depths = sorted({r["depth"] for r in subset})
        width = 0.35
        for i, res in enumerate([False, True]):
            vals = []
            for d in depths:
                cand = [r for r in subset if r["depth"] == d and r["residual"] == res]
                vals.append(max(max(r["val_acc"]) for r in cand) if cand else 0)
            xs = [d + (i - 0.5) * width for d in depths]
            ax.bar(xs, vals, width, label="residual" if res else "plain",
                   color="tab:red" if res else "tab:blue", alpha=0.8)
        ax.set_xticks(depths)
        ax.set_title(title)
        ax.set_xlabel("depth (blocks)")
        ax.set_ylabel("best val acc")
        ax.set_ylim(0, 1.05)
        ax.legend(fontsize=8)
        ax.grid(alpha=0.3, axis="y")
    fig.tight_layout()
    fig.savefig(OUT / fname, dpi=150)
    plt.close(fig)


def main():
    runs = load_runs()
    plot_curves(runs, bn=True, metric="val_acc", ylabel="val accuracy",
                fname="val_acc_bn.png")
    plot_curves(runs, bn=True, metric="train_loss", ylabel="train loss",
                fname="train_loss_bn.png", logy=True)
    plot_curves(runs, bn=False, metric="val_acc", ylabel="val accuracy",
                fname="val_acc_nobn.png")
    plot_curves(runs, bn=False, metric="train_loss", ylabel="train loss",
                fname="train_loss_nobn.png", logy=True)
    plot_summary(runs, "summary.png")
    print(f"saved plots to {OUT}")


if __name__ == "__main__":
    main()