Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Plotly. Landing page.
Browse files- app.py +94 -3
- requirements.txt +2 -1
- src/display/css_html_js.py +4 -0
app.py
CHANGED
@@ -2,6 +2,7 @@
|
|
2 |
|
3 |
import gradio as gr
|
4 |
import pandas as pd
|
|
|
5 |
from apscheduler.schedulers.background import BackgroundScheduler
|
6 |
from gradio_leaderboard import Leaderboard, SelectColumns
|
7 |
from huggingface_hub import whoami
|
@@ -205,6 +206,69 @@ def _select_example_tab(choice: str):
|
|
205 |
)
|
206 |
|
207 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
208 |
# Force light theme even if HF user prefers dark
|
209 |
blocks = gr.Blocks(
|
210 |
css=custom_css,
|
@@ -214,7 +278,34 @@ blocks = gr.Blocks(
|
|
214 |
with blocks:
|
215 |
|
216 |
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
217 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
218 |
|
219 |
gr.Image(
|
220 |
"assets/banner.png",
|
@@ -336,7 +427,7 @@ with blocks:
|
|
336 |
gr.HTML(WHAT_IS_F1_HTML_AFTER_TIER1FIG_TAIL)
|
337 |
|
338 |
# Rename tab to "Leaderboard" and cap at 800px width
|
339 |
-
with gr.TabItem("Leaderboard", elem_id="formulaone-leaderboard-tab-table", id=
|
340 |
gr.Markdown(
|
341 |
"""
|
342 |
Welcome to the FormulaOne leaderboard. This table tracks the performance of various systems on the FormulaOne benchmark.
|
@@ -348,7 +439,7 @@ with blocks:
|
|
348 |
assert leaderboard_df is not None
|
349 |
leaderboard_component = init_leaderboard(leaderboard_df)
|
350 |
|
351 |
-
with gr.TabItem("Submit Solutions", elem_id="formulaone-submit-tab-table", id=
|
352 |
logger.info("Tab submission")
|
353 |
with gr.Column():
|
354 |
with gr.Row():
|
|
|
2 |
|
3 |
import gradio as gr
|
4 |
import pandas as pd
|
5 |
+
import plotly.graph_objects as go # NEW: for interactive chart
|
6 |
from apscheduler.schedulers.background import BackgroundScheduler
|
7 |
from gradio_leaderboard import Leaderboard, SelectColumns
|
8 |
from huggingface_hub import whoami
|
|
|
206 |
)
|
207 |
|
208 |
|
209 |
+
# === Static, made-up results for the landing chart (not tied to leaderboard) ===
|
210 |
+
TIER_TOTALS = {"Warmup": 100, "Tier 1": 100, "Tier 2": 20} # dataset sizes
|
211 |
+
MODELS_ORDER = ["GPT-5", "Gemini 2.5 Pro", "Grok 4", "Claude Opus 4", "o3 Pro"]
|
212 |
+
STATIC_RESULTS = {
|
213 |
+
"Warmup": {
|
214 |
+
"GPT-5": 95,
|
215 |
+
"Gemini 2.5 Pro": 90,
|
216 |
+
"Grok 4": 84,
|
217 |
+
"Claude Opus 4": 92,
|
218 |
+
"o3 Pro": 88,
|
219 |
+
},
|
220 |
+
"Tier 1": {
|
221 |
+
"GPT-5": 38,
|
222 |
+
"Gemini 2.5 Pro": 30,
|
223 |
+
"Grok 4": 24,
|
224 |
+
"Claude Opus 4": 35,
|
225 |
+
"o3 Pro": 28,
|
226 |
+
},
|
227 |
+
"Tier 2": {
|
228 |
+
"GPT-5": 1,
|
229 |
+
"Gemini 2.5 Pro": 0,
|
230 |
+
"Grok 4": 0,
|
231 |
+
"Claude Opus 4": 0,
|
232 |
+
"o3 Pro": 0,
|
233 |
+
},
|
234 |
+
}
|
235 |
+
|
236 |
+
|
237 |
+
def build_accuracy_figure(tier: str):
|
238 |
+
"""Create interactive bar chart with hover showing 'solved / total'."""
|
239 |
+
results = STATIC_RESULTS.get(tier, {})
|
240 |
+
total = TIER_TOTALS[tier]
|
241 |
+
x = MODELS_ORDER
|
242 |
+
y = [results[m] for m in x]
|
243 |
+
hover = [f"{m}<br><b>{v}/{total}</b> problems solved" for m, v in zip(x, y)]
|
244 |
+
|
245 |
+
fig = go.Figure(
|
246 |
+
data=[
|
247 |
+
go.Bar(
|
248 |
+
x=x,
|
249 |
+
y=y,
|
250 |
+
text=[f"{v}/{total}" for v in y],
|
251 |
+
textposition="auto",
|
252 |
+
hovertext=hover,
|
253 |
+
hoverinfo="text",
|
254 |
+
marker_line_width=0.5,
|
255 |
+
)
|
256 |
+
]
|
257 |
+
)
|
258 |
+
fig.update_layout(
|
259 |
+
template="plotly_white",
|
260 |
+
margin=dict(l=30, r=20, t=10, b=40),
|
261 |
+
yaxis=dict(title="# Problems Solved", range=[0, total], dtick=max(5, total // 10)),
|
262 |
+
xaxis=dict(title=None),
|
263 |
+
height=420,
|
264 |
+
)
|
265 |
+
return fig
|
266 |
+
|
267 |
+
|
268 |
+
# Precompute initial figure (Warmup)
|
269 |
+
_initial_accuracy_fig = build_accuracy_figure("Warmup")
|
270 |
+
|
271 |
+
|
272 |
# Force light theme even if HF user prefers dark
|
273 |
blocks = gr.Blocks(
|
274 |
css=custom_css,
|
|
|
278 |
with blocks:
|
279 |
|
280 |
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
281 |
+
# === NEW LANDING TAB (first) ===
|
282 |
+
with gr.TabItem("Model Accuracy on FormulaOne", id=0, elem_id="landing-accuracy-tab"):
|
283 |
+
gr.Markdown(
|
284 |
+
"The chart below summarizes static (non-live) results for model performance on FormulaOne.",
|
285 |
+
elem_classes="markdown-text",
|
286 |
+
)
|
287 |
+
|
288 |
+
# Selector aligned to the top-right (see CSS)
|
289 |
+
with gr.Row(elem_id="f1-tier-select-row"):
|
290 |
+
tier_selector = gr.Radio(
|
291 |
+
choices=list(TIER_TOTALS.keys()),
|
292 |
+
value="Warmup",
|
293 |
+
label=None,
|
294 |
+
show_label=False,
|
295 |
+
elem_id="f1-tier-select",
|
296 |
+
)
|
297 |
+
|
298 |
+
accuracy_plot = gr.Plot(value=_initial_accuracy_fig)
|
299 |
+
|
300 |
+
# Wire selector → plot
|
301 |
+
tier_selector.change(
|
302 |
+
lambda t: build_accuracy_figure(t),
|
303 |
+
inputs=tier_selector,
|
304 |
+
outputs=accuracy_plot,
|
305 |
+
)
|
306 |
+
|
307 |
+
# Existing "What is FormulaOne" tab
|
308 |
+
with gr.TabItem("What is FormulaOne", id=1, elem_id="what-is-tab"):
|
309 |
|
310 |
gr.Image(
|
311 |
"assets/banner.png",
|
|
|
427 |
gr.HTML(WHAT_IS_F1_HTML_AFTER_TIER1FIG_TAIL)
|
428 |
|
429 |
# Rename tab to "Leaderboard" and cap at 800px width
|
430 |
+
with gr.TabItem("Leaderboard", elem_id="formulaone-leaderboard-tab-table", id=2):
|
431 |
gr.Markdown(
|
432 |
"""
|
433 |
Welcome to the FormulaOne leaderboard. This table tracks the performance of various systems on the FormulaOne benchmark.
|
|
|
439 |
assert leaderboard_df is not None
|
440 |
leaderboard_component = init_leaderboard(leaderboard_df)
|
441 |
|
442 |
+
with gr.TabItem("Submit Solutions", elem_id="formulaone-submit-tab-table", id=3):
|
443 |
logger.info("Tab submission")
|
444 |
with gr.Column():
|
445 |
with gr.Row():
|
requirements.txt
CHANGED
@@ -14,4 +14,5 @@ python-dateutil
|
|
14 |
tqdm
|
15 |
transformers
|
16 |
tokenizers>=0.15.0
|
17 |
-
sentencepiece
|
|
|
|
14 |
tqdm
|
15 |
transformers
|
16 |
tokenizers>=0.15.0
|
17 |
+
sentencepiece
|
18 |
+
plotly>=5
|
src/display/css_html_js.py
CHANGED
@@ -18,6 +18,10 @@ custom_css = """
|
|
18 |
/* requested 710px */
|
19 |
#f1-examples { max-width: 710px; margin: 0 auto; }
|
20 |
|
|
|
|
|
|
|
|
|
21 |
/* Text */
|
22 |
.f1-p, .f1-li { line-height: 1.75; color: #374151; text-wrap: pretty; overflow-wrap: break-word; hyphens: auto; }
|
23 |
|
|
|
18 |
/* requested 710px */
|
19 |
#f1-examples { max-width: 710px; margin: 0 auto; }
|
20 |
|
21 |
+
/* NEW: landing tab width + tier selector alignment */
|
22 |
+
#landing-accuracy-tab { max-width: 800px; margin-left: auto; margin-right: auto; }
|
23 |
+
#f1-tier-select-row { justify-content: flex-end; margin-bottom: 6px; }
|
24 |
+
|
25 |
/* Text */
|
26 |
.f1-p, .f1-li { line-height: 1.75; color: #374151; text-wrap: pretty; overflow-wrap: break-word; hyphens: auto; }
|
27 |
|