File size: 5,011 Bytes
0c0ff0e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
129
130
131
132
133
134
135
136
137
138
139
140
"""Plot k-step detour metrics vs k for the architectures on one figure per metric.

Data is loaded automatically from the kdetour result files saved by
maze_kstep_detour_test.py (out/maze_kdetour/kdetour_*.npz, key 'table' with
column names in 'columns'). You only need to set TASK / DATASET / configs
(via the command line, see --help). One PNG is produced per metric in
METRICS, written to out/plot/.
"""
import os
import argparse
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt

# ---------------------------------------------------------------------------
# Defaults (overridable on the command line, see --help).
# ---------------------------------------------------------------------------
TASK = 'I1'
DATASET = '10M'

# Model configs: transformer-family uses TF_CONFIG, RNN/SSM-family uses RNN_CONFIG.
TF_CONFIG = '6_6_384'
RNN_CONFIG = '12_384'

# kdetour result settings (match how the .npz files were produced).
CKPT_ITER = 10000
PATH_TYPE = 'RWs'
KDETOUR_DIR = 'out/maze_kdetour'
OUT_DIR = 'out/plot'

# (display name, file model key, is_transformer_family). Configs come from args.
MODEL_SPECS = [
    ('Transformer',          'transformer',         True),
    ('Transformer-NextLat',  'transformer_nextlat', True),
    ('Mamba',                'mamba',               False),
    ('Mamba-2',              'mamba2',              False),
    ('Gated-DeltaNet',       'gated_deltanet',      False),
    ('GRU',                  'gru',                 False),
]

MARKERS = ['o', 's', '^', 'D', 'P', 'v']

# Each metric maps to a column name in the kdetour table.
# ylim=None means autoscale (used for JSD which is not a percentage).
METRICS = [
    ('probe_acc',   'probe accuracy (%)', (-2, 105)),
    ('c_match_jsd', 'c_match_jsd',        None),
    ('reach_acc',   'reach accuracy (%)', (-2, 105)),
]


def parse_args():
    p = argparse.ArgumentParser(description='Plot k-step detour metrics from kdetour npz files.')
    p.add_argument('--task', default=TASK)
    p.add_argument('--dataset', default=DATASET)
    p.add_argument('--tf_config', default=TF_CONFIG)
    p.add_argument('--rnn_config', default=RNN_CONFIG)
    p.add_argument('--ckpt_iter', type=int, default=CKPT_ITER)
    p.add_argument('--path_type', default=PATH_TYPE)
    p.add_argument('--kdetour_dir', default=KDETOUR_DIR)
    p.add_argument('--out_dir', default=OUT_DIR)
    return p.parse_args()


def load_table(args, model_key, config):
    """Return (k array, {col_name: values}) from the kdetour npz, or (None, None)."""
    fname = f'kdetour_{args.task}_{args.path_type}_{args.ckpt_iter}_{args.dataset}_{model_key}_{config}.npz'
    path = os.path.join(args.kdetour_dir, fname)
    if not os.path.exists(path):
        print(f"[missing] {path}")
        return None, None
    d = np.load(path, allow_pickle=True)
    if 'table' not in d or 'columns' not in d:
        print(f"[no table] {path}")
        return None, None
    table = np.asarray(d['table'], dtype=float)
    cols = [str(c) for c in d['columns']]
    by_col = {c: table[:, i] for i, c in enumerate(cols)}
    k = by_col.get('k')
    return k, by_col


def plot_metric(args, loaded, metric, ylabel, ylim):
    plt.figure(figsize=(9, 5.5))
    all_k = None
    plotted = 0
    for (name, key, is_tf), marker in zip(MODEL_SPECS, MARKERS):
        k, by_col = loaded[key]
        if k is None or metric not in by_col:
            continue
        config = args.tf_config if is_tf else args.rnn_config
        plt.plot(k, by_col[metric], marker=marker, markersize=6, linewidth=2,
                 label=f'{name} ({config})')
        if all_k is None or len(k) > len(all_k):
            all_k = k
        plotted += 1

    if plotted == 0:
        print(f"[{metric}] nothing to plot: no kdetour npz files found.")
        plt.close()
        return

    plt.xlabel('detour length k (steps)', fontsize=12)
    plt.ylabel(ylabel, fontsize=12)
    plt.title(f'k-step detour {metric} (Task {args.task}, {args.dataset})', fontsize=13)
    ax = plt.gca()
    ticks = [int(x) for x in all_k]
    ax.set_xticks(ticks)
    ax.set_xticklabels([str(t) for t in ticks])
    for t, lbl in zip(ticks, ax.get_xticklabels()):
        if t >= 100:
            lbl.set_color('red')
    if ylim is not None:
        plt.ylim(*ylim)
    plt.grid(True, alpha=0.3)
    plt.legend(fontsize=10, framealpha=0.9)
    plt.tight_layout()
    out = os.path.join(
        args.out_dir,
        f'kstep_{metric}_{args.task}_{args.dataset}_{args.tf_config}_{args.rnn_config}.png')
    plt.savefig(out, dpi=150)
    plt.close()
    print(f"Saved figure to {out}")


def main():
    args = parse_args()
    os.makedirs(args.out_dir, exist_ok=True)
    loaded = {}
    for name, key, is_tf in MODEL_SPECS:
        config = args.tf_config if is_tf else args.rnn_config
        loaded[key] = load_table(args, key, config)
    for metric, ylabel, ylim in METRICS:
        plot_metric(args, loaded, metric, ylabel, ylim)


if __name__ == '__main__':
    main()