Spaces:
Running
Running
Empty PR for testing
#9
by
gneubig - opened
- app.py +4 -27
- main_page.py +41 -81
- ui_components.py +2 -2
app.py
CHANGED
|
@@ -379,37 +379,14 @@ with demo.route("About", "/about"):
|
|
| 379 |
logger.info("All routes configured")
|
| 380 |
|
| 381 |
# Mount the REST API on /api
|
| 382 |
-
from fastapi import FastAPI
|
| 383 |
-
from fastapi.responses import RedirectResponse
|
| 384 |
-
from starlette.middleware.base import BaseHTTPMiddleware
|
| 385 |
from api import api_app
|
| 386 |
|
| 387 |
-
|
| 388 |
-
|
| 389 |
-
"""Middleware to redirect root path "/" to "/home".
|
| 390 |
-
|
| 391 |
-
This fixes the 307 trailing slash redirect issue (Gradio bug #11071) that
|
| 392 |
-
occurs when Gradio is mounted at "/" - FastAPI's default behavior redirects
|
| 393 |
-
"/" to "//", which breaks routing on HuggingFace Spaces.
|
| 394 |
-
|
| 395 |
-
See: https://github.com/gradio-app/gradio/issues/11071
|
| 396 |
-
"""
|
| 397 |
-
async def dispatch(self, request: Request, call_next):
|
| 398 |
-
if request.url.path == "/":
|
| 399 |
-
return RedirectResponse(url="/home", status_code=302)
|
| 400 |
-
return await call_next(request)
|
| 401 |
-
|
| 402 |
-
|
| 403 |
-
# Create a parent FastAPI app with redirect_slashes=False to prevent
|
| 404 |
-
# automatic trailing slash redirects that cause issues with Gradio
|
| 405 |
-
root_app = FastAPI(redirect_slashes=False)
|
| 406 |
-
|
| 407 |
-
# Add middleware to handle root path redirect to /home
|
| 408 |
-
root_app.add_middleware(RootRedirectMiddleware)
|
| 409 |
-
|
| 410 |
root_app.mount("/api", api_app)
|
| 411 |
|
| 412 |
-
# Mount Gradio app
|
| 413 |
app = gr.mount_gradio_app(root_app, demo, path="/")
|
| 414 |
logger.info("REST API mounted at /api, Gradio app mounted at /")
|
| 415 |
|
|
|
|
| 379 |
logger.info("All routes configured")
|
| 380 |
|
| 381 |
# Mount the REST API on /api
|
| 382 |
+
from fastapi import FastAPI
|
|
|
|
|
|
|
| 383 |
from api import api_app
|
| 384 |
|
| 385 |
+
# Create a parent FastAPI app that will host both the API and Gradio
|
| 386 |
+
root_app = FastAPI()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 387 |
root_app.mount("/api", api_app)
|
| 388 |
|
| 389 |
+
# Mount Gradio app - root redirect is handled by the proxy
|
| 390 |
app = gr.mount_gradio_app(root_app, demo, path="/")
|
| 391 |
logger.info("REST API mounted at /api, Gradio app mounted at /")
|
| 392 |
|
main_page.py
CHANGED
|
@@ -1,7 +1,6 @@
|
|
| 1 |
import matplotlib
|
| 2 |
matplotlib.use('Agg')
|
| 3 |
import gradio as gr
|
| 4 |
-
import pandas as pd
|
| 5 |
|
| 6 |
|
| 7 |
from ui_components import (
|
|
@@ -27,32 +26,6 @@ from constants import MARK_BY_DEFAULT
|
|
| 27 |
CACHED_VIEWERS = {}
|
| 28 |
CACHED_TAG_MAPS = {}
|
| 29 |
|
| 30 |
-
|
| 31 |
-
def filter_complete_entries(df: pd.DataFrame) -> pd.DataFrame:
|
| 32 |
-
if df.empty:
|
| 33 |
-
return df.copy()
|
| 34 |
-
|
| 35 |
-
category_score_columns = [
|
| 36 |
-
'Issue Resolution Score',
|
| 37 |
-
'Frontend Score',
|
| 38 |
-
'Greenfield Score',
|
| 39 |
-
'Testing Score',
|
| 40 |
-
'Information Gathering Score',
|
| 41 |
-
]
|
| 42 |
-
|
| 43 |
-
if all(column in df.columns for column in category_score_columns):
|
| 44 |
-
return df[df[category_score_columns].notna().all(axis=1)].copy()
|
| 45 |
-
|
| 46 |
-
if 'Categories Completed' in df.columns:
|
| 47 |
-
categories_completed = pd.to_numeric(df['Categories Completed'], errors='coerce')
|
| 48 |
-
return df[categories_completed >= 5].copy()
|
| 49 |
-
|
| 50 |
-
if 'Categories Attempted' in df.columns:
|
| 51 |
-
return df[df['Categories Attempted'] == '5/5'].copy()
|
| 52 |
-
|
| 53 |
-
return df.copy()
|
| 54 |
-
|
| 55 |
-
|
| 56 |
def build_page():
|
| 57 |
with gr.Row(elem_id="intro-row"):
|
| 58 |
with gr.Column(scale=1):
|
|
@@ -65,91 +38,78 @@ def build_page():
|
|
| 65 |
|
| 66 |
test_df, test_tag_map = get_full_leaderboard_data("test")
|
| 67 |
if not test_df.empty:
|
| 68 |
-
|
|
|
|
| 69 |
full_df=test_df,
|
| 70 |
tag_map=test_tag_map,
|
| 71 |
category_name=CATEGORY_NAME,
|
| 72 |
split_name="test"
|
| 73 |
)
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
has_complete_entries = len(test_df_complete) > 0
|
| 77 |
-
|
| 78 |
if 'Openness' in test_df.columns:
|
| 79 |
test_df_open = test_df[test_df['Openness'].str.lower() == 'open'].copy()
|
| 80 |
else:
|
| 81 |
test_df_open = test_df.copy()
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
initial_df = test_df_complete if has_complete_entries else test_df
|
| 85 |
-
|
| 86 |
# --- Winners by Category Section ---
|
| 87 |
gr.Markdown("---")
|
| 88 |
gr.HTML('<h2>Winners by Category</h2>', elem_id="winners-header")
|
| 89 |
gr.Markdown("Top 5 performing systems in each benchmark category.")
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
|
|
|
| 96 |
# --- New Visualization Sections ---
|
| 97 |
gr.Markdown("---")
|
| 98 |
-
|
| 99 |
# Evolution Over Time Section
|
| 100 |
gr.HTML('<h2>Evolution Over Time</h2>', elem_id="evolution-header")
|
| 101 |
gr.Markdown("Track how model performance has improved over time based on release dates.")
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
)
|
| 107 |
-
|
| 108 |
gr.Markdown("---")
|
| 109 |
-
|
| 110 |
# Open Model Accuracy by Size Section (always shows open models only by design)
|
| 111 |
gr.HTML('<h2>Open Model Accuracy by Size</h2>', elem_id="size-accuracy-header")
|
| 112 |
gr.Markdown("Compare open-weights model performance against their parameter count.")
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
evolution_fig = create_evolution_over_time_chart(
|
| 127 |
-
size_fig = create_accuracy_by_size_chart(
|
| 128 |
-
|
| 129 |
return winners_html, evolution_fig, size_fig
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
show_open_only_input = show_open_only_checkbox if show_open_only_checkbox is not None else gr.State(value=False)
|
| 133 |
-
extra_section_inputs = [show_incomplete_input, show_open_only_input, mark_by_dropdown]
|
| 134 |
-
|
| 135 |
-
if show_incomplete_checkbox is not None:
|
| 136 |
-
show_incomplete_checkbox.change(
|
| 137 |
-
fn=update_extra_sections,
|
| 138 |
-
inputs=extra_section_inputs,
|
| 139 |
-
outputs=[winners_component, evolution_component, size_component]
|
| 140 |
-
)
|
| 141 |
-
|
| 142 |
if show_open_only_checkbox is not None:
|
| 143 |
show_open_only_checkbox.change(
|
| 144 |
fn=update_extra_sections,
|
| 145 |
-
inputs=
|
| 146 |
outputs=[winners_component, evolution_component, size_component]
|
| 147 |
)
|
| 148 |
-
|
| 149 |
if mark_by_dropdown is not None:
|
| 150 |
mark_by_dropdown.change(
|
| 151 |
fn=update_extra_sections,
|
| 152 |
-
inputs=
|
| 153 |
outputs=[winners_component, evolution_component, size_component]
|
| 154 |
)
|
| 155 |
|
|
|
|
| 1 |
import matplotlib
|
| 2 |
matplotlib.use('Agg')
|
| 3 |
import gradio as gr
|
|
|
|
| 4 |
|
| 5 |
|
| 6 |
from ui_components import (
|
|
|
|
| 26 |
CACHED_VIEWERS = {}
|
| 27 |
CACHED_TAG_MAPS = {}
|
| 28 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
def build_page():
|
| 30 |
with gr.Row(elem_id="intro-row"):
|
| 31 |
with gr.Column(scale=1):
|
|
|
|
| 38 |
|
| 39 |
test_df, test_tag_map = get_full_leaderboard_data("test")
|
| 40 |
if not test_df.empty:
|
| 41 |
+
# Get the checkbox and dropdown returned from create_leaderboard_display
|
| 42 |
+
show_open_only_checkbox, mark_by_dropdown = create_leaderboard_display(
|
| 43 |
full_df=test_df,
|
| 44 |
tag_map=test_tag_map,
|
| 45 |
category_name=CATEGORY_NAME,
|
| 46 |
split_name="test"
|
| 47 |
)
|
| 48 |
+
|
| 49 |
+
# Prepare open-only filtered dataframe for Winners and Evolution
|
|
|
|
|
|
|
| 50 |
if 'Openness' in test_df.columns:
|
| 51 |
test_df_open = test_df[test_df['Openness'].str.lower() == 'open'].copy()
|
| 52 |
else:
|
| 53 |
test_df_open = test_df.copy()
|
| 54 |
+
|
|
|
|
|
|
|
|
|
|
| 55 |
# --- Winners by Category Section ---
|
| 56 |
gr.Markdown("---")
|
| 57 |
gr.HTML('<h2>Winners by Category</h2>', elem_id="winners-header")
|
| 58 |
gr.Markdown("Top 5 performing systems in each benchmark category.")
|
| 59 |
+
|
| 60 |
+
# Create both all and open-only versions of winners HTML
|
| 61 |
+
winners_html_all = create_winners_by_category_html(test_df, top_n=5)
|
| 62 |
+
winners_html_open = create_winners_by_category_html(test_df_open, top_n=5)
|
| 63 |
+
|
| 64 |
+
winners_component = gr.HTML(winners_html_all, elem_id="winners-by-category")
|
| 65 |
+
|
| 66 |
# --- New Visualization Sections ---
|
| 67 |
gr.Markdown("---")
|
| 68 |
+
|
| 69 |
# Evolution Over Time Section
|
| 70 |
gr.HTML('<h2>Evolution Over Time</h2>', elem_id="evolution-header")
|
| 71 |
gr.Markdown("Track how model performance has improved over time based on release dates.")
|
| 72 |
+
|
| 73 |
+
# Create initial evolution chart with default mark_by
|
| 74 |
+
evolution_fig_all = create_evolution_over_time_chart(test_df, MARK_BY_DEFAULT)
|
| 75 |
+
|
| 76 |
+
evolution_component = gr.Plot(value=evolution_fig_all, elem_id="evolution-chart")
|
| 77 |
+
|
| 78 |
gr.Markdown("---")
|
| 79 |
+
|
| 80 |
# Open Model Accuracy by Size Section (always shows open models only by design)
|
| 81 |
gr.HTML('<h2>Open Model Accuracy by Size</h2>', elem_id="size-accuracy-header")
|
| 82 |
gr.Markdown("Compare open-weights model performance against their parameter count.")
|
| 83 |
+
|
| 84 |
+
size_fig = create_accuracy_by_size_chart(test_df, MARK_BY_DEFAULT)
|
| 85 |
+
size_component = gr.Plot(value=size_fig, elem_id="size-accuracy-chart")
|
| 86 |
+
|
| 87 |
+
# Update function for Winners, Evolution, and Size charts based on filters
|
| 88 |
+
def update_extra_sections(show_open_only, mark_by):
|
| 89 |
+
# Select the appropriate dataframe based on open_only filter
|
| 90 |
+
df_to_use = test_df_open if show_open_only else test_df
|
| 91 |
+
|
| 92 |
+
# Winners HTML (not affected by mark_by, only open_only)
|
| 93 |
+
winners_html = winners_html_open if show_open_only else winners_html_all
|
| 94 |
+
|
| 95 |
+
# Regenerate charts with current mark_by setting
|
| 96 |
+
evolution_fig = create_evolution_over_time_chart(df_to_use, mark_by)
|
| 97 |
+
size_fig = create_accuracy_by_size_chart(test_df, mark_by) # Size chart always uses full df (filters internally)
|
| 98 |
+
|
| 99 |
return winners_html, evolution_fig, size_fig
|
| 100 |
+
|
| 101 |
+
# Connect both checkbox and dropdown to update all extra sections
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 102 |
if show_open_only_checkbox is not None:
|
| 103 |
show_open_only_checkbox.change(
|
| 104 |
fn=update_extra_sections,
|
| 105 |
+
inputs=[show_open_only_checkbox, mark_by_dropdown],
|
| 106 |
outputs=[winners_component, evolution_component, size_component]
|
| 107 |
)
|
| 108 |
+
|
| 109 |
if mark_by_dropdown is not None:
|
| 110 |
mark_by_dropdown.change(
|
| 111 |
fn=update_extra_sections,
|
| 112 |
+
inputs=[show_open_only_checkbox if show_open_only_checkbox else gr.State(value=False), mark_by_dropdown],
|
| 113 |
outputs=[winners_component, evolution_component, size_component]
|
| 114 |
)
|
| 115 |
|
ui_components.py
CHANGED
|
@@ -1032,8 +1032,8 @@ def create_leaderboard_display(
|
|
| 1032 |
outputs=[dataframe_component, cost_plot_component, runtime_plot_component]
|
| 1033 |
)
|
| 1034 |
|
| 1035 |
-
# Return the
|
| 1036 |
-
return
|
| 1037 |
|
| 1038 |
# # --- Detailed Benchmark Display ---
|
| 1039 |
def create_benchmark_details_display(
|
|
|
|
| 1032 |
outputs=[dataframe_component, cost_plot_component, runtime_plot_component]
|
| 1033 |
)
|
| 1034 |
|
| 1035 |
+
# Return the show_open_only_checkbox and mark_by_dropdown so they can be used to update other sections
|
| 1036 |
+
return show_open_only_checkbox, mark_by_dropdown
|
| 1037 |
|
| 1038 |
# # --- Detailed Benchmark Display ---
|
| 1039 |
def create_benchmark_details_display(
|