File size: 6,773 Bytes
e8c37dc 23299dd e8c37dc 9c0900a e8c37dc 9c0900a ddc171b e8c37dc ddc171b e8c37dc 9c0900a e8c37dc 9c0900a e8c37dc 9c0900a e8c37dc 23299dd e8c37dc 23299dd e8c37dc 23299dd e8c37dc 0e019ea e8c37dc 23299dd e8c37dc 23299dd e8c37dc 23299dd e8c37dc f152c96 e8c37dc 23299dd e8c37dc 23299dd e8c37dc 23299dd f152c96 e8c37dc | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 | #!/usr/bin/env python3
"""Plot Laguna Martini pruning insights from published artifacts."""
from __future__ import annotations
import argparse
import json
from pathlib import Path
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser()
parser.add_argument("--summary-path", type=Path, default=Path("results/summary.json"))
parser.add_argument(
"--mask-path",
type=Path,
default=Path("artifacts/group_keep_mask_25pct.npy"),
)
parser.add_argument(
"--scores-path",
type=Path,
default=Path("artifacts/atomic_scores.npy"),
)
parser.add_argument("--output-dir", type=Path, default=Path("assets"))
return parser.parse_args()
def configure_style() -> None:
sns.set_theme(style="whitegrid", context="talk")
plt.rcParams.update(
{
"figure.facecolor": "#081522",
"axes.facecolor": "#102235",
"axes.edgecolor": "#7ed9e6",
"axes.labelcolor": "#e6fbff",
"axes.titlecolor": "#e6fbff",
"grid.color": "#34536a",
"text.color": "#e6fbff",
"xtick.color": "#cceff4",
"ytick.color": "#cceff4",
}
)
def plot_pruning_curve(summary: dict, output_dir: Path) -> None:
rows = summary["native_group_loss_sweep"]
baseline = summary["native_group_loss_sweep_baseline"]["perplexity"]
x = np.array([row["groups_pruned_ratio"] * 100 for row in rows])
y = np.array([(row["perplexity"] / baseline - 1) * 100 for row in rows])
random_controls = [
summary["random_native_group_10pct_control"],
summary["random_native_group_25pct_control"],
]
random_x = np.array([0, *[row["groups_pruned_ratio"] * 100 for row in random_controls]])
random_y = np.array([0, *[(row["perplexity"] / baseline - 1) * 100 for row in random_controls]])
fig, ax = plt.subplots(figsize=(10, 5.6))
sns.lineplot(
x=x,
y=y,
marker="o",
linewidth=3,
markersize=10,
color="#78dce8",
label="HEAPr-style importance ranking",
ax=ax,
)
sns.lineplot(
x=random_x,
y=random_y,
marker="o",
linewidth=2.5,
markersize=9,
linestyle="--",
color="#f5a65b",
label="Random control",
ax=ax,
)
selected = int(np.flatnonzero(x == 10)[0])
ax.scatter([x[selected]], [y[selected]], s=180, color="#b7d36b", zorder=4)
ax.annotate(
"Recommended operating point\n10% pruning, +1.66% perplexity",
xy=(x[selected], y[selected]),
xytext=(48, 34),
textcoords="offset points",
color="#e6fbff",
fontsize=12,
arrowprops={"arrowstyle": "->", "color": "#b7d36b"},
)
ax.set(
title="Importance Ranking Outperforms Random Pruning",
xlabel="Structured MoE parameters identified for removal (%)",
ylabel="Perplexity increase vs. BF16 baseline (%)",
)
ax.set_xlim(-1, 42)
ax.set_ylim(-1, 44)
ax.legend(frameon=False)
sns.despine(ax=ax)
fig.tight_layout()
fig.savefig(output_dir / "pruning_quality_curve.png", dpi=180)
plt.close(fig)
def plot_atomic_importance_by_layer(scores: np.ndarray, output_dir: Path) -> None:
positive_scores = scores[scores > 0]
score_floor = max(float(positive_scores.min()), np.finfo(scores.dtype).tiny)
log_scores = np.log10(np.clip(scores, score_floor, None))
layers = np.arange(scores.shape[0])
medians = np.median(log_scores, axis=(1, 2))
fig, ax = plt.subplots(figsize=(12, 5.8))
sns.boxplot(
data=[layer.reshape(-1) for layer in log_scores],
color="#78dce8",
fliersize=0,
linewidth=0.8,
width=0.72,
ax=ax,
)
sns.lineplot(
x=layers,
y=medians,
color="#b7d36b",
linewidth=2.5,
marker="o",
markersize=4,
label="Layer median",
ax=ax,
)
ax.set(
title="Later Layers Have Lower-Importance Atomic Experts",
xlabel="Sparse layer index",
ylabel="log10(atomic importance score)",
)
ax.set_ylim(np.quantile(log_scores, 0.005), np.quantile(log_scores, 0.995))
ax.tick_params(axis="x", labelsize=9)
ax.legend(frameon=False)
sns.despine(ax=ax)
fig.tight_layout()
fig.savefig(output_dir / "atomic_importance_by_layer.png", dpi=180)
plt.close(fig)
def plot_gsm8k_curve(summary: dict, output_dir: Path) -> None:
gsm8k = summary["evaluation"]["gsm8k_cot_8shot_first_10pct"]
x = np.array([0, 10, 20, 25])
strict = np.array(
[
gsm8k["baseline_strict_exact_match"],
gsm8k["native_group_10pct_strict_exact_match"],
gsm8k["native_group_20pct_strict_exact_match"],
gsm8k["native_group_25pct_strict_exact_match"],
]
)
flexible = np.array(
[
gsm8k["baseline_flexible_exact_match"],
gsm8k["native_group_10pct_flexible_exact_match"],
gsm8k["native_group_20pct_flexible_exact_match"],
gsm8k["native_group_25pct_flexible_exact_match"],
]
)
fig, ax = plt.subplots(figsize=(10, 5.6))
sns.lineplot(
x=x,
y=strict,
marker="o",
linewidth=3,
markersize=10,
color="#78dce8",
label="Strict exact match",
ax=ax,
)
sns.lineplot(
x=x,
y=flexible,
marker="o",
linewidth=3,
markersize=10,
color="#b7d36b",
label="Flexible exact match",
ax=ax,
)
ax.set(
title="GSM8K-CoT Is Increasingly Sensitive to Pruning",
xlabel="Structured MoE parameters identified for removal (%)",
ylabel="Exact match on paired first-10% subset",
)
ax.set_xticks(x)
ax.set_ylim(0.62, 0.98)
ax.legend(frameon=False)
sns.despine(ax=ax)
fig.tight_layout()
fig.savefig(output_dir / "gsm8k_pruning_curve.png", dpi=180)
plt.close(fig)
def main() -> None:
args = parse_args()
args.output_dir.mkdir(parents=True, exist_ok=True)
summary = json.loads(args.summary_path.read_text())
mask = np.load(args.mask_path)
scores = np.load(args.scores_path)
if mask.shape != (39, 256, 8):
raise ValueError(f"unexpected mask shape: {mask.shape}")
if scores.shape != (39, 256, 512):
raise ValueError(f"unexpected scores shape: {scores.shape}")
configure_style()
plot_pruning_curve(summary, args.output_dir)
plot_atomic_importance_by_layer(scores, args.output_dir)
plot_gsm8k_curve(summary, args.output_dir)
if __name__ == "__main__":
main()
|