File size: 6,167 Bytes
8ce2c21 | 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 | #!/usr/bin/env python3
"""Render public benchmark assets for TensorMind 1.5 Preview."""
from __future__ import annotations
import json
from pathlib import Path
import matplotlib as mpl
import matplotlib.pyplot as plt
import numpy as np
from matplotlib.offsetbox import AnnotationBbox, OffsetImage
ROOT = Path(__file__).resolve().parent
DATA = json.loads((ROOT / "benchmark-results.json").read_text())
LOGO = ROOT / "tensorplay-ai-logo.png"
BG = "#07111F"
PANEL = "#0C1A2C"
PANEL_ALT = "#10243A"
WHITE = "#F4F8FF"
MUTED = "#8EA5C3"
GRID = "#263A52"
CYAN = "#2ED7FF"
BLUE = "#267BFF"
ORANGE = "#FF9D42"
mpl.rcParams.update(
{
"font.family": "DejaVu Sans",
"axes.facecolor": BG,
"figure.facecolor": BG,
"savefig.facecolor": BG,
"text.color": WHITE,
"axes.labelcolor": MUTED,
"xtick.color": MUTED,
"ytick.color": MUTED,
"axes.edgecolor": GRID,
"svg.fonttype": "none",
}
)
def add_logo(fig: plt.Figure) -> None:
rgba = plt.imread(LOGO)
alpha = rgba[..., 3].copy() if rgba.shape[-1] == 4 else 1.0 - rgba[..., :3].mean(axis=-1)
white_logo = np.ones((*alpha.shape, 4), dtype=float)
white_logo[..., 3] = alpha
fig.add_artist(
AnnotationBbox(
OffsetImage(white_logo, zoom=0.105),
(0.84, 0.927),
xycoords="figure fraction",
frameon=False,
)
)
def header(fig: plt.Figure, title: str, subtitle: str) -> None:
fig.text(0.055, 0.905, "TensorMind 1.5 Preview", fontsize=13, color=CYAN, weight="bold")
fig.text(0.055, 0.843, title, fontsize=28, weight="bold")
fig.text(0.055, 0.792, subtitle, fontsize=11, color=MUTED)
add_logo(fig)
def footer(fig: plt.Figure) -> None:
fig.text(
0.055,
0.055,
"Protocol lm-eval 0.4.12 路 SGLang 0.5.14 路 0-shot 路 full datasets 路 batch 48 路 fixed seeds",
fontsize=8.5,
color=MUTED,
)
fig.text(0.955, 0.055, "Accuracy, higher is better", ha="right", fontsize=8.5, color=MUTED)
def save(fig: plt.Figure, name: str) -> None:
fig.savefig(ROOT / f"{name}.png", dpi=200)
fig.savefig(ROOT / f"{name}.svg")
plt.close(fig)
def render_suite() -> None:
results = DATA["results"]
metrics = [
("CMMLU", results["cmmlu"]),
("AGIEval-CN", results["agieval_cn"]),
("A-CLUE", results["a_clue"]),
("C-Eval", results["c_eval"]),
("TMMLU+", results["tmmlu_plus"]),
]
fig = plt.figure(figsize=(14, 7.875), dpi=200)
header(fig, "Chinese benchmark suite", "Five full-dataset evaluations under one matched zero-shot protocol")
gs = fig.add_gridspec(1, 12, left=0.095, right=0.955, top=0.72, bottom=0.13, wspace=1.2)
ax = fig.add_subplot(gs[0, :8])
ax.set_facecolor(PANEL)
for spine in ax.spines.values():
spine.set_visible(False)
names = [name for name, _ in metrics][::-1]
values = [value for _, value in metrics][::-1]
y = np.arange(len(metrics))
ax.barh(y, values, height=0.46, color=[BLUE, CYAN, BLUE, CYAN, BLUE], alpha=0.95)
ax.set_xlim(0, 35)
ax.set_yticks(y, names, fontsize=10.5)
ax.set_xticks([0, 10, 20, 30])
ax.tick_params(axis="both", length=0, pad=10)
ax.grid(axis="x", color=GRID, linewidth=0.8, alpha=0.75)
ax.set_axisbelow(True)
for yi, value in enumerate(values):
ax.text(value + 0.45, yi, f"{value:.4f}", va="center", fontsize=10, color=WHITE, weight="bold")
ax.set_xlabel("Accuracy (%)", loc="right", fontsize=9, labelpad=10)
ax_card = fig.add_subplot(gs[0, 9:])
ax_card.set_facecolor(PANEL_ALT)
ax_card.set_xticks([])
ax_card.set_yticks([])
for spine in ax_card.spines.values():
spine.set_visible(False)
ax_card.text(0.10, 0.86, "FIVE-SUITE MACRO", fontsize=8.5, color=CYAN, weight="bold", transform=ax_card.transAxes)
ax_card.text(0.10, 0.64, f"{results['five_suite_macro']:.4f}", fontsize=36, color=WHITE, weight="bold", transform=ax_card.transAxes)
ax_card.text(0.10, 0.53, "full-dataset accuracy", fontsize=9.5, color=MUTED, transform=ax_card.transAxes)
ax_card.plot([0.10, 0.90], [0.43, 0.43], color=GRID, linewidth=1.0, transform=ax_card.transAxes)
ax_card.text(0.10, 0.32, "5", fontsize=19, color=CYAN, weight="bold", transform=ax_card.transAxes)
ax_card.text(0.21, 0.33, "benchmark suites", fontsize=9.5, color=MUTED, transform=ax_card.transAxes)
ax_card.text(0.10, 0.18, "0-shot", fontsize=19, color=ORANGE, weight="bold", transform=ax_card.transAxes)
ax_card.text(0.52, 0.19, "matched protocol", fontsize=9.5, color=MUTED, transform=ax_card.transAxes)
footer(fig)
save(fig, "benchmark-suite")
def render_scorecard() -> None:
results = DATA["results"]
cards = [
("CMMLU", results["cmmlu"], BLUE),
("AGIEval-CN", results["agieval_cn"], CYAN),
("A-CLUE", results["a_clue"], BLUE),
("C-Eval", results["c_eval"], ORANGE),
("TMMLU+", results["tmmlu_plus"], CYAN),
("5-suite macro", results["five_suite_macro"], WHITE),
]
fig = plt.figure(figsize=(14, 7.875), dpi=200)
header(fig, "Benchmark scorecard", "TensorMind 1.5 Preview 路 full-dataset accuracy (%)")
gs = fig.add_gridspec(2, 3, left=0.08, right=0.92, top=0.70, bottom=0.18, wspace=0.10, hspace=0.14)
for idx, (name, value, color) in enumerate(cards):
ax = fig.add_subplot(gs[idx // 3, idx % 3])
ax.set_facecolor(PANEL_ALT if idx == 5 else PANEL)
ax.set_xticks([])
ax.set_yticks([])
for spine in ax.spines.values():
spine.set_visible(False)
ax.add_patch(plt.Rectangle((0.0, 0.0), 0.018, 1.0, color=color, transform=ax.transAxes, lw=0))
ax.text(0.09, 0.70, name.upper(), fontsize=9, color=MUTED, weight="bold", transform=ax.transAxes)
ax.text(0.09, 0.29, f"{value:.4f}", fontsize=27, color=color, weight="bold", transform=ax.transAxes)
ax.text(0.09, 0.12, "accuracy", fontsize=8.5, color=MUTED, transform=ax.transAxes)
footer(fig)
save(fig, "benchmark-matrix")
if __name__ == "__main__":
render_suite()
render_scorecard()
|