geographdb-core 0.5.4

Geometric graph database core - 3D spatial indexing for code analysis
Documentation 
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

#!/usr/bin/env python3
"""Plot a generate_confidence_routing.csv run.

Shows:
  - generated token ids for the dense baseline and the target mode,
    with mismatches highlighted.
  - confidence per generated position (only meaningful for --mode gated).
"""

import sys
from pathlib import Path

import matplotlib.pyplot as plt
import numpy as np


def main() -> None:
    if len(sys.argv) < 2:
        csv_path = Path("target/generate_confidence_routing.csv")
    else:
        csv_path = Path(sys.argv[1])

    if not csv_path.exists():
        raise SystemExit(f"CSV not found: {csv_path}")

    data = np.genfromtxt(csv_path, delimiter=",", names=True, dtype=None, encoding=None)
    pos = data["position"]
    dense = data["dense_token"]
    target = data["target_token"]
    matched = data["matched"]
    routed = data["routed_low_rank"]
    conf = data["confidence"]

    fig, axes = plt.subplots(2, 1, figsize=(12, 7), sharex=True)

    ax = axes[0]
    ax.plot(pos, dense, "-", label="dense", color="tab:blue", alpha=0.7)
    ax.plot(pos, target, "-", label="target", color="tab:orange", alpha=0.7)
    mismatch = pos[matched == 0]
    ax.scatter(
        mismatch,
        target[matched == 0],
        color="tab:red",
        s=30,
        zorder=5,
        label="mismatch",
    )
    ax.set_ylabel("token id")
    ax.set_title(f"{csv_path.name} — agreement {matched.mean()*100:.1f}%")
    ax.legend()
    ax.grid(True, ls="--", alpha=0.4)

    ax = axes[1]
    routed_mask = routed == 1
    if np.any(routed_mask):
        ax.bar(pos[routed_mask], conf[routed_mask], color="tab:green", width=0.8, label="low-rank")
    fallback_mask = (~routed_mask) & np.isfinite(conf)
    if np.any(fallback_mask):
        ax.bar(
            pos[fallback_mask],
            conf[fallback_mask],
            color="tab:red",
            width=0.8,
            label="dense fallback",
        )
    ax.axhline(np.nanmean(conf), color="black", ls="--", lw=1, label="mean confidence")
    ax.set_xlabel("generated position")
    ax.set_ylabel("confidence")
    ax.set_title("per-position confidence (gated mode only)")
    ax.legend()
    ax.grid(True, ls="--", alpha=0.4)

    out = csv_path.with_suffix(".png")
    fig.tight_layout()
    fig.savefig(out, dpi=150)
    print(f"Saved {out}")


if __name__ == "__main__":
    main()