Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Pratik Bhavsar
commited on
Commit
·
7d1a2ad
1
Parent(s):
91da2cc
improved colors
Browse files- tabs/leaderboard.py +8 -0
- utils.py +28 -6
- visualization.py +65 -44
tabs/leaderboard.py
CHANGED
|
@@ -39,6 +39,10 @@ def filter_leaderboard(df, model_type, category, sort_by):
|
|
| 39 |
--hover-bg: #2d2e32;
|
| 40 |
--note-bg: #2d2e32;
|
| 41 |
--note-text: #a1a1aa;
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
}}
|
| 43 |
}}
|
| 44 |
|
|
@@ -50,6 +54,10 @@ def filter_leaderboard(df, model_type, category, sort_by):
|
|
| 50 |
--hover-bg: #f3f4f6;
|
| 51 |
--note-bg: #f3f4f6;
|
| 52 |
--note-text: #4b5563;
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
}}
|
| 54 |
}}
|
| 55 |
|
|
|
|
| 39 |
--hover-bg: #2d2e32;
|
| 40 |
--note-bg: #2d2e32;
|
| 41 |
--note-text: #a1a1aa;
|
| 42 |
+
--accent-blue: #60A5FA;
|
| 43 |
+
--accent-purple: #A78BFA;
|
| 44 |
+
--accent-pink: #F472B6;
|
| 45 |
+
--score-bg: rgba(255, 255, 255, 0.1);
|
| 46 |
}}
|
| 47 |
}}
|
| 48 |
|
|
|
|
| 54 |
--hover-bg: #f3f4f6;
|
| 55 |
--note-bg: #f3f4f6;
|
| 56 |
--note-text: #4b5563;
|
| 57 |
+
--accent-blue: #3B82F6;
|
| 58 |
+
--accent-purple: #8B5CF6;
|
| 59 |
+
--accent-pink: #EC4899;
|
| 60 |
+
--score-bg: rgba(0, 0, 0, 0.1);
|
| 61 |
}}
|
| 62 |
}}
|
| 63 |
|
utils.py
CHANGED
|
@@ -1,3 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
def get_rank_badge(rank):
|
| 2 |
"""Generate HTML for rank badge with appropriate styling"""
|
| 3 |
badge_styles = {
|
|
@@ -41,7 +61,8 @@ def get_rank_badge(rank):
|
|
| 41 |
|
| 42 |
def get_type_badge(model_type):
|
| 43 |
"""Generate HTML for model type badge"""
|
| 44 |
-
colors =
|
|
|
|
| 45 |
bg_color = colors.get(model_type, "#4F46E5")
|
| 46 |
return f"""
|
| 47 |
<div style="
|
|
@@ -60,23 +81,24 @@ def get_type_badge(model_type):
|
|
| 60 |
|
| 61 |
|
| 62 |
def get_score_bar(score):
|
| 63 |
-
"""Generate HTML for score bar"""
|
| 64 |
width = score * 100
|
| 65 |
return f"""
|
| 66 |
<div style="display: flex; align-items: center; gap: 12px; width: 100%;">
|
| 67 |
<div style="
|
| 68 |
flex-grow: 1;
|
| 69 |
-
height:
|
| 70 |
background: var(--score-bg, rgba(255, 255, 255, 0.1));
|
| 71 |
-
border-radius:
|
| 72 |
overflow: hidden;
|
| 73 |
max-width: 200px;
|
| 74 |
">
|
| 75 |
<div style="
|
| 76 |
width: {width}%;
|
| 77 |
height: 100%;
|
| 78 |
-
background: var(--accent-
|
| 79 |
-
border-radius:
|
|
|
|
| 80 |
"></div>
|
| 81 |
</div>
|
| 82 |
<span style="
|
|
|
|
| 1 |
+
def get_chart_colors():
|
| 2 |
+
# if is_dark_theme():
|
| 3 |
+
# return {
|
| 4 |
+
# "Private": "#60A5FA", # accent-blue
|
| 5 |
+
# "Open source": "#A78BFA", # accent-purple
|
| 6 |
+
# "performance_bands": ["#DCFCE7", "#FEF9C3", "#FEE2E2"],
|
| 7 |
+
# "text": "#FFFFFF",
|
| 8 |
+
# "background": "#1a1b1e",
|
| 9 |
+
# "grid": (1, 1, 1, 0.1), # RGBA tuple for grid
|
| 10 |
+
# }
|
| 11 |
+
return {
|
| 12 |
+
"Private": "#3F78FA", # accent-blue light
|
| 13 |
+
"Open source": "#A13AE2", # accent-purple light
|
| 14 |
+
"performance_bands": ["#DCFCE7", "#FEF9C3", "#FEE2E2"],
|
| 15 |
+
"text": "#111827",
|
| 16 |
+
"background": "#FFFFFF",
|
| 17 |
+
"grid": (0, 0, 0, 0.1), # RGBA tuple for grid
|
| 18 |
+
}
|
| 19 |
+
|
| 20 |
+
|
| 21 |
def get_rank_badge(rank):
|
| 22 |
"""Generate HTML for rank badge with appropriate styling"""
|
| 23 |
badge_styles = {
|
|
|
|
| 61 |
|
| 62 |
def get_type_badge(model_type):
|
| 63 |
"""Generate HTML for model type badge"""
|
| 64 |
+
colors = get_chart_colors()
|
| 65 |
+
colors = {"Private": colors["Private"], "Open source": colors["Open source"]}
|
| 66 |
bg_color = colors.get(model_type, "#4F46E5")
|
| 67 |
return f"""
|
| 68 |
<div style="
|
|
|
|
| 81 |
|
| 82 |
|
| 83 |
def get_score_bar(score):
|
| 84 |
+
"""Generate HTML for score bar with gradient styling"""
|
| 85 |
width = score * 100
|
| 86 |
return f"""
|
| 87 |
<div style="display: flex; align-items: center; gap: 12px; width: 100%;">
|
| 88 |
<div style="
|
| 89 |
flex-grow: 1;
|
| 90 |
+
height: 8px;
|
| 91 |
background: var(--score-bg, rgba(255, 255, 255, 0.1));
|
| 92 |
+
border-radius: 4px;
|
| 93 |
overflow: hidden;
|
| 94 |
max-width: 200px;
|
| 95 |
">
|
| 96 |
<div style="
|
| 97 |
width: {width}%;
|
| 98 |
height: 100%;
|
| 99 |
+
background: linear-gradient(90deg, var(--accent-blue, #60A5FA), var(--accent-purple, #A78BFA));
|
| 100 |
+
border-radius: 4px;
|
| 101 |
+
transition: width 0.3s ease;
|
| 102 |
"></div>
|
| 103 |
</div>
|
| 104 |
<span style="
|
visualization.py
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
|
|
| 1 |
import matplotlib
|
| 2 |
import matplotlib.pyplot as plt
|
| 3 |
import numpy as np
|
|
@@ -5,54 +6,68 @@ import plotly.graph_objects as go
|
|
| 5 |
|
| 6 |
|
| 7 |
def setup_matplotlib():
|
| 8 |
-
"""Set up matplotlib configuration."""
|
| 9 |
matplotlib.use("Agg")
|
| 10 |
plt.close("all")
|
| 11 |
|
| 12 |
|
| 13 |
def get_performance_chart(df, category_name="Overall"):
|
| 14 |
plt.close("all")
|
|
|
|
| 15 |
score_column = "Category Score"
|
| 16 |
df_sorted = df.sort_values(score_column, ascending=True)
|
| 17 |
-
colors = {"Private": "#4F46E5", "Open source": "#16A34A"}
|
| 18 |
|
| 19 |
height = max(8, len(df_sorted) * 0.8)
|
| 20 |
fig, ax = plt.subplots(figsize=(16, height))
|
| 21 |
plt.rcParams.update({"font.size": 12})
|
| 22 |
|
|
|
|
|
|
|
|
|
|
| 23 |
try:
|
| 24 |
bars = ax.barh(
|
| 25 |
np.arange(len(df_sorted)),
|
| 26 |
df_sorted[score_column],
|
| 27 |
-
height=0.
|
|
|
|
| 28 |
color=[colors[t] for t in df_sorted["Model Type"]],
|
| 29 |
)
|
| 30 |
|
| 31 |
ax.set_title(
|
| 32 |
-
f"Model Performance
|
| 33 |
pad=20,
|
| 34 |
fontsize=20,
|
| 35 |
fontweight="bold",
|
|
|
|
| 36 |
)
|
| 37 |
-
ax.set_xlabel("Average Score", fontsize=14, labelpad=10)
|
| 38 |
ax.set_xlim(0.0, 1.0)
|
| 39 |
|
| 40 |
ax.set_yticks(np.arange(len(df_sorted)))
|
| 41 |
-
ax.set_yticklabels(df_sorted["Model"], fontsize=12)
|
| 42 |
|
| 43 |
plt.subplots_adjust(left=0.35)
|
| 44 |
|
| 45 |
for i, v in enumerate(df_sorted[score_column]):
|
| 46 |
ax.text(
|
| 47 |
-
v + 0.01,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
)
|
| 49 |
|
| 50 |
-
ax.grid(True, axis="x", linestyle="--", alpha=0.2)
|
| 51 |
ax.spines[["top", "right"]].set_visible(False)
|
|
|
|
|
|
|
| 52 |
|
| 53 |
legend_elements = [
|
| 54 |
plt.Rectangle((0, 0), 1, 1, facecolor=color, label=label)
|
| 55 |
-
for label, color in
|
|
|
|
|
|
|
| 56 |
]
|
| 57 |
ax.legend(
|
| 58 |
handles=legend_elements,
|
|
@@ -60,6 +75,8 @@ def get_performance_chart(df, category_name="Overall"):
|
|
| 60 |
loc="lower right",
|
| 61 |
fontsize=12,
|
| 62 |
title_fontsize=14,
|
|
|
|
|
|
|
| 63 |
)
|
| 64 |
|
| 65 |
plt.tight_layout()
|
|
@@ -67,7 +84,6 @@ def get_performance_chart(df, category_name="Overall"):
|
|
| 67 |
finally:
|
| 68 |
plt.close(fig)
|
| 69 |
|
| 70 |
-
|
| 71 |
def create_radar_plot(df, model_names):
|
| 72 |
datasets = [col for col in df.columns[7:] if col != "IO Cost"]
|
| 73 |
fig = go.Figure()
|
|
@@ -132,26 +148,24 @@ def create_radar_plot(df, model_names):
|
|
| 132 |
|
| 133 |
|
| 134 |
def get_performance_cost_chart(df, category_name="Overall"):
|
| 135 |
-
|
| 136 |
fig, ax = plt.subplots(figsize=(12, 8), dpi=300)
|
| 137 |
|
| 138 |
-
|
| 139 |
-
ax.
|
| 140 |
-
ax.
|
| 141 |
-
fig.patch.set_facecolor("white")
|
| 142 |
-
|
| 143 |
-
colors = {"Private": "#4F46E5", "Open source": "#16A34A"}
|
| 144 |
-
performance_colors = ["#DCFCE7", "#FEF9C3", "#FEE2E2"]
|
| 145 |
|
| 146 |
score_column = "Category Score"
|
| 147 |
|
| 148 |
-
# Plot data points
|
| 149 |
for _, row in df.iterrows():
|
| 150 |
color = colors[row["Model Type"]]
|
| 151 |
size = 100 if row[score_column] > 0.85 else 80
|
| 152 |
-
edge_color =
|
|
|
|
|
|
|
|
|
|
|
|
|
| 153 |
|
| 154 |
-
# Plot scatter points
|
| 155 |
ax.scatter(
|
| 156 |
row["IO Cost"],
|
| 157 |
row[score_column] * 100,
|
|
@@ -160,11 +174,12 @@ def get_performance_cost_chart(df, category_name="Overall"):
|
|
| 160 |
alpha=0.9,
|
| 161 |
edgecolor=edge_color,
|
| 162 |
linewidth=1,
|
| 163 |
-
zorder=5,
|
| 164 |
)
|
| 165 |
|
| 166 |
-
|
| 167 |
-
|
|
|
|
| 168 |
|
| 169 |
ax.annotate(
|
| 170 |
f"{row['Model']}\n(${row['IO Cost']:.2f})",
|
|
@@ -172,50 +187,56 @@ def get_performance_cost_chart(df, category_name="Overall"):
|
|
| 172 |
xytext=(5, 5),
|
| 173 |
textcoords="offset points",
|
| 174 |
fontsize=8,
|
|
|
|
| 175 |
bbox=bbox_props,
|
| 176 |
zorder=6,
|
| 177 |
)
|
| 178 |
|
| 179 |
-
# Configure axes
|
| 180 |
ax.set_xscale("log")
|
| 181 |
-
ax.set_xlim(0.08, 40)
|
| 182 |
ax.set_ylim(60, 95)
|
| 183 |
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 187 |
|
| 188 |
-
# Add legend
|
| 189 |
legend_elements = [
|
| 190 |
-
plt.scatter([], [], c=
|
| 191 |
-
for label,
|
| 192 |
]
|
| 193 |
ax.legend(
|
| 194 |
handles=legend_elements,
|
| 195 |
loc="upper right",
|
| 196 |
frameon=True,
|
| 197 |
-
facecolor="
|
| 198 |
edgecolor="none",
|
| 199 |
fontsize=9,
|
|
|
|
| 200 |
)
|
| 201 |
|
| 202 |
-
# Set title
|
| 203 |
ax.set_title(
|
| 204 |
-
f"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 205 |
)
|
| 206 |
|
| 207 |
-
|
| 208 |
-
for y1, y2, color in zip([85, 75, 60], [95, 85, 75], performance_colors):
|
| 209 |
ax.axhspan(y1, y2, alpha=0.2, color=color, zorder=1)
|
| 210 |
|
| 211 |
-
|
| 212 |
-
ax.tick_params(axis="both", which="
|
| 213 |
-
ax.tick_params(axis="both", which="minor", labelsize=8)
|
| 214 |
-
|
| 215 |
-
# Add minor ticks for log scale
|
| 216 |
ax.xaxis.set_minor_locator(plt.LogLocator(base=10.0, subs=np.arange(2, 10) * 0.1))
|
| 217 |
|
| 218 |
-
|
| 219 |
-
|
| 220 |
|
|
|
|
| 221 |
return fig
|
|
|
|
| 1 |
+
from utils import get_chart_colors
|
| 2 |
import matplotlib
|
| 3 |
import matplotlib.pyplot as plt
|
| 4 |
import numpy as np
|
|
|
|
| 6 |
|
| 7 |
|
| 8 |
def setup_matplotlib():
|
|
|
|
| 9 |
matplotlib.use("Agg")
|
| 10 |
plt.close("all")
|
| 11 |
|
| 12 |
|
| 13 |
def get_performance_chart(df, category_name="Overall"):
|
| 14 |
plt.close("all")
|
| 15 |
+
colors = get_chart_colors()
|
| 16 |
score_column = "Category Score"
|
| 17 |
df_sorted = df.sort_values(score_column, ascending=True)
|
|
|
|
| 18 |
|
| 19 |
height = max(8, len(df_sorted) * 0.8)
|
| 20 |
fig, ax = plt.subplots(figsize=(16, height))
|
| 21 |
plt.rcParams.update({"font.size": 12})
|
| 22 |
|
| 23 |
+
fig.patch.set_facecolor(colors["background"])
|
| 24 |
+
ax.set_facecolor(colors["background"])
|
| 25 |
+
|
| 26 |
try:
|
| 27 |
bars = ax.barh(
|
| 28 |
np.arange(len(df_sorted)),
|
| 29 |
df_sorted[score_column],
|
| 30 |
+
height=0.4,
|
| 31 |
+
capstyle="round",
|
| 32 |
color=[colors[t] for t in df_sorted["Model Type"]],
|
| 33 |
)
|
| 34 |
|
| 35 |
ax.set_title(
|
| 36 |
+
f"Model Performance - {category_name}",
|
| 37 |
pad=20,
|
| 38 |
fontsize=20,
|
| 39 |
fontweight="bold",
|
| 40 |
+
color=colors["text"],
|
| 41 |
)
|
| 42 |
+
ax.set_xlabel("Average Score", fontsize=14, labelpad=10, color=colors["text"])
|
| 43 |
ax.set_xlim(0.0, 1.0)
|
| 44 |
|
| 45 |
ax.set_yticks(np.arange(len(df_sorted)))
|
| 46 |
+
ax.set_yticklabels(df_sorted["Model"], fontsize=12, color=colors["text"])
|
| 47 |
|
| 48 |
plt.subplots_adjust(left=0.35)
|
| 49 |
|
| 50 |
for i, v in enumerate(df_sorted[score_column]):
|
| 51 |
ax.text(
|
| 52 |
+
v + 0.01,
|
| 53 |
+
i,
|
| 54 |
+
f"{v:.3f}",
|
| 55 |
+
va="center",
|
| 56 |
+
fontsize=12,
|
| 57 |
+
fontweight="bold",
|
| 58 |
+
color=colors["text"],
|
| 59 |
)
|
| 60 |
|
| 61 |
+
ax.grid(True, axis="x", linestyle="--", alpha=0.2, color=colors["grid"])
|
| 62 |
ax.spines[["top", "right"]].set_visible(False)
|
| 63 |
+
ax.spines[["bottom", "left"]].set_color(colors["grid"])
|
| 64 |
+
ax.tick_params(colors=colors["text"])
|
| 65 |
|
| 66 |
legend_elements = [
|
| 67 |
plt.Rectangle((0, 0), 1, 1, facecolor=color, label=label)
|
| 68 |
+
for label, color in {
|
| 69 |
+
k: colors[k] for k in ["Private", "Open source"]
|
| 70 |
+
}.items()
|
| 71 |
]
|
| 72 |
ax.legend(
|
| 73 |
handles=legend_elements,
|
|
|
|
| 75 |
loc="lower right",
|
| 76 |
fontsize=12,
|
| 77 |
title_fontsize=14,
|
| 78 |
+
facecolor=colors["background"],
|
| 79 |
+
labelcolor=colors["text"],
|
| 80 |
)
|
| 81 |
|
| 82 |
plt.tight_layout()
|
|
|
|
| 84 |
finally:
|
| 85 |
plt.close(fig)
|
| 86 |
|
|
|
|
| 87 |
def create_radar_plot(df, model_names):
|
| 88 |
datasets = [col for col in df.columns[7:] if col != "IO Cost"]
|
| 89 |
fig = go.Figure()
|
|
|
|
| 148 |
|
| 149 |
|
| 150 |
def get_performance_cost_chart(df, category_name="Overall"):
|
| 151 |
+
colors = get_chart_colors()
|
| 152 |
fig, ax = plt.subplots(figsize=(12, 8), dpi=300)
|
| 153 |
|
| 154 |
+
fig.patch.set_facecolor(colors["background"])
|
| 155 |
+
ax.set_facecolor(colors["background"])
|
| 156 |
+
ax.grid(True, linestyle="--", alpha=0.15, which="both", color=colors["grid"])
|
|
|
|
|
|
|
|
|
|
|
|
|
| 157 |
|
| 158 |
score_column = "Category Score"
|
| 159 |
|
|
|
|
| 160 |
for _, row in df.iterrows():
|
| 161 |
color = colors[row["Model Type"]]
|
| 162 |
size = 100 if row[score_column] > 0.85 else 80
|
| 163 |
+
edge_color = (
|
| 164 |
+
colors["Private"]
|
| 165 |
+
if row["Model Type"] == "Private"
|
| 166 |
+
else colors["Open source"]
|
| 167 |
+
)
|
| 168 |
|
|
|
|
| 169 |
ax.scatter(
|
| 170 |
row["IO Cost"],
|
| 171 |
row[score_column] * 100,
|
|
|
|
| 174 |
alpha=0.9,
|
| 175 |
edgecolor=edge_color,
|
| 176 |
linewidth=1,
|
| 177 |
+
zorder=5,
|
| 178 |
)
|
| 179 |
|
| 180 |
+
bbox_props = dict(
|
| 181 |
+
boxstyle="round,pad=0.3", fc=colors["background"], ec="none", alpha=0.8
|
| 182 |
+
)
|
| 183 |
|
| 184 |
ax.annotate(
|
| 185 |
f"{row['Model']}\n(${row['IO Cost']:.2f})",
|
|
|
|
| 187 |
xytext=(5, 5),
|
| 188 |
textcoords="offset points",
|
| 189 |
fontsize=8,
|
| 190 |
+
color=colors["text"],
|
| 191 |
bbox=bbox_props,
|
| 192 |
zorder=6,
|
| 193 |
)
|
| 194 |
|
|
|
|
| 195 |
ax.set_xscale("log")
|
| 196 |
+
ax.set_xlim(0.08, 40)
|
| 197 |
ax.set_ylim(60, 95)
|
| 198 |
|
| 199 |
+
ax.set_xlabel(
|
| 200 |
+
"I/O Cost per Million Tokens ($)",
|
| 201 |
+
fontsize=10,
|
| 202 |
+
labelpad=10,
|
| 203 |
+
color=colors["text"],
|
| 204 |
+
)
|
| 205 |
+
ax.set_ylabel(
|
| 206 |
+
"Model Performance Score", fontsize=10, labelpad=10, color=colors["text"]
|
| 207 |
+
)
|
| 208 |
|
|
|
|
| 209 |
legend_elements = [
|
| 210 |
+
plt.scatter([], [], c=colors[label], label=label, s=80)
|
| 211 |
+
for label in ["Private", "Open source"]
|
| 212 |
]
|
| 213 |
ax.legend(
|
| 214 |
handles=legend_elements,
|
| 215 |
loc="upper right",
|
| 216 |
frameon=True,
|
| 217 |
+
facecolor=colors["background"],
|
| 218 |
edgecolor="none",
|
| 219 |
fontsize=9,
|
| 220 |
+
labelcolor=colors["text"],
|
| 221 |
)
|
| 222 |
|
|
|
|
| 223 |
ax.set_title(
|
| 224 |
+
f"Performance vs. Cost - {category_name}",
|
| 225 |
+
fontsize=12,
|
| 226 |
+
pad=15,
|
| 227 |
+
fontweight="bold",
|
| 228 |
+
color=colors["text"],
|
| 229 |
)
|
| 230 |
|
| 231 |
+
for y1, y2, color in zip([85, 75, 60], [95, 85, 75], colors["performance_bands"]):
|
|
|
|
| 232 |
ax.axhspan(y1, y2, alpha=0.2, color=color, zorder=1)
|
| 233 |
|
| 234 |
+
ax.tick_params(axis="both", which="major", labelsize=9, colors=colors["text"])
|
| 235 |
+
ax.tick_params(axis="both", which="minor", labelsize=8, colors=colors["text"])
|
|
|
|
|
|
|
|
|
|
| 236 |
ax.xaxis.set_minor_locator(plt.LogLocator(base=10.0, subs=np.arange(2, 10) * 0.1))
|
| 237 |
|
| 238 |
+
for spine in ax.spines.values():
|
| 239 |
+
spine.set_color(colors["grid"])
|
| 240 |
|
| 241 |
+
plt.tight_layout()
|
| 242 |
return fig
|