Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Separate leaderboard section
Browse files
app.py
CHANGED
@@ -211,187 +211,191 @@ def filter_models(
|
|
211 |
|
212 |
leaderboard_df = filter_models(leaderboard_df, [t.to_str(" : ") for t in ModelType], list(NUMERIC_INTERVALS.keys()), [i.value.name for i in Precision], [i.value.name for i in AddSpecialTokens], [i.value.name for i in NumFewShots], False, False, False)
|
213 |
|
214 |
-
with gr.Blocks(
|
215 |
-
gr.
|
216 |
-
|
217 |
-
|
218 |
-
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
219 |
-
with gr.TabItem("🏅 LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
|
220 |
with gr.Row():
|
221 |
-
|
222 |
-
with
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
240 |
-
|
241 |
-
|
242 |
-
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
|
256 |
-
|
257 |
-
|
258 |
-
|
259 |
-
|
260 |
-
interactive=True,
|
261 |
-
elem_id="filter-columns-type",
|
262 |
-
)
|
263 |
-
filter_columns_precision = gr.CheckboxGroup(
|
264 |
-
label="Precision",
|
265 |
-
choices=[i.value.name for i in Precision],
|
266 |
-
value=[i.value.name for i in Precision],
|
267 |
-
interactive=True,
|
268 |
-
elem_id="filter-columns-precision",
|
269 |
-
)
|
270 |
-
filter_columns_size = gr.CheckboxGroup(
|
271 |
-
label="Model sizes (in billions of parameters)",
|
272 |
-
choices=list(NUMERIC_INTERVALS.keys()),
|
273 |
-
value=list(NUMERIC_INTERVALS.keys()),
|
274 |
-
interactive=True,
|
275 |
-
elem_id="filter-columns-size",
|
276 |
-
)
|
277 |
-
filter_columns_add_special_tokens = gr.CheckboxGroup(
|
278 |
-
label="Add Special Tokens",
|
279 |
-
choices=[i.value.name for i in AddSpecialTokens],
|
280 |
-
value=[i.value.name for i in AddSpecialTokens],
|
281 |
-
interactive=True,
|
282 |
-
elem_id="filter-columns-add-special-tokens",
|
283 |
-
)
|
284 |
-
filter_columns_num_few_shots = gr.CheckboxGroup(
|
285 |
-
label="Num Few Shots",
|
286 |
-
choices=[i.value.name for i in NumFewShots],
|
287 |
-
value=[i.value.name for i in NumFewShots],
|
288 |
-
interactive=True,
|
289 |
-
elem_id="filter-columns-num-few-shots",
|
290 |
-
)
|
291 |
-
|
292 |
-
leaderboard_df_filtered = filter_models(leaderboard_df, [t.to_str(" : ") for t in ModelType], list(NUMERIC_INTERVALS.keys()), [i.value.name for i in Precision], [i.value.name for i in AddSpecialTokens], [i.value.name for i in NumFewShots], False, False, False)
|
293 |
-
# initial_columns = [c.name for c in fields(AutoEvalColumn) if c.never_hidden or c.displayed_by_default]
|
294 |
-
# leaderboard_df_filtered = select_columns(leaderboard_df_filtered, initial_columns)
|
295 |
-
|
296 |
-
# leaderboard_table = gr.components.Dataframe(
|
297 |
-
# value=leaderboard_df_filtered,
|
298 |
-
# headers=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + shown_columns.value,
|
299 |
-
# datatype=TYPES,
|
300 |
-
# elem_id="leaderboard-table",
|
301 |
-
# interactive=False,
|
302 |
-
# visible=True,
|
303 |
-
# )
|
304 |
-
|
305 |
-
# DataFrameの初期化部分のみを修正
|
306 |
-
initial_columns = ['T'] + [c.name for c in fields(AutoEvalColumn) if (c.never_hidden or c.displayed_by_default) and c.name != 'T']
|
307 |
-
leaderboard_df_filtered = select_columns(leaderboard_df, initial_columns)
|
308 |
-
|
309 |
-
# Model列のリンク形式を修正
|
310 |
-
leaderboard_df_filtered['Model'] = leaderboard_df_filtered['Model'].apply(
|
311 |
-
lambda x: f'[{x.split(">")[-2].split("<")[0]}]({x.split("href=")[1].split(chr(34))[1]})' if isinstance(x, str) and 'href=' in x else x
|
312 |
)
|
313 |
-
|
314 |
-
|
315 |
-
|
316 |
-
|
317 |
-
|
318 |
-
|
319 |
-
# DataFrameコンポーネントの初期化
|
320 |
-
leaderboard_table = gr.components.Dataframe(
|
321 |
-
value=leaderboard_df_filtered,
|
322 |
-
headers=initial_columns,
|
323 |
-
datatype=TYPES,
|
324 |
-
elem_id="leaderboard-table",
|
325 |
-
interactive=False,
|
326 |
-
visible=True
|
327 |
)
|
328 |
-
|
329 |
-
|
330 |
-
|
331 |
-
value=
|
332 |
-
|
333 |
-
|
334 |
-
visible=False,
|
335 |
)
|
336 |
-
|
337 |
-
|
338 |
-
[
|
339 |
-
|
340 |
-
|
341 |
-
|
342 |
-
filter_columns_precision,
|
343 |
-
filter_columns_size,
|
344 |
-
filter_columns_add_special_tokens,
|
345 |
-
filter_columns_num_few_shots,
|
346 |
-
deleted_models_visibility,
|
347 |
-
merged_models_visibility,
|
348 |
-
flagged_models_visibility,
|
349 |
-
search_bar,
|
350 |
-
],
|
351 |
-
leaderboard_table,
|
352 |
)
|
353 |
-
|
354 |
-
|
355 |
-
|
356 |
-
|
357 |
-
|
358 |
-
|
359 |
-
hidden_leaderboard_table_for_search,
|
360 |
-
shown_columns,
|
361 |
-
filter_columns_type,
|
362 |
-
filter_columns_precision,
|
363 |
-
filter_columns_size,
|
364 |
-
filter_columns_add_special_tokens,
|
365 |
-
filter_columns_num_few_shots,
|
366 |
-
deleted_models_visibility,
|
367 |
-
merged_models_visibility,
|
368 |
-
flagged_models_visibility,
|
369 |
-
search_bar,
|
370 |
-
],
|
371 |
-
leaderboard_table,
|
372 |
)
|
373 |
-
|
374 |
-
|
375 |
-
|
376 |
-
|
377 |
-
|
378 |
-
|
379 |
-
|
380 |
-
|
381 |
-
|
382 |
-
|
383 |
-
|
384 |
-
|
385 |
-
|
386 |
-
|
387 |
-
|
388 |
-
|
389 |
-
|
390 |
-
|
391 |
-
|
392 |
-
|
393 |
-
|
394 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
395 |
|
396 |
with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=2):
|
397 |
gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
|
|
|
211 |
|
212 |
leaderboard_df = filter_models(leaderboard_df, [t.to_str(" : ") for t in ModelType], list(NUMERIC_INTERVALS.keys()), [i.value.name for i in Precision], [i.value.name for i in AddSpecialTokens], [i.value.name for i in NumFewShots], False, False, False)
|
213 |
|
214 |
+
with gr.Blocks() as demo_leaderboard:
|
215 |
+
with gr.Row():
|
216 |
+
with gr.Column():
|
|
|
|
|
|
|
217 |
with gr.Row():
|
218 |
+
search_bar = gr.Textbox(
|
219 |
+
placeholder=" 🔍 Search for your model (separate multiple queries with `;`) and press ENTER...",
|
220 |
+
show_label=False,
|
221 |
+
elem_id="search-bar",
|
222 |
+
)
|
223 |
+
with gr.Row():
|
224 |
+
shown_columns = gr.CheckboxGroup(
|
225 |
+
choices=[
|
226 |
+
c.name
|
227 |
+
for c in fields(AutoEvalColumn)
|
228 |
+
if not c.hidden and not c.never_hidden# and not c.dummy
|
229 |
+
],
|
230 |
+
value=[
|
231 |
+
c.name
|
232 |
+
for c in fields(AutoEvalColumn)
|
233 |
+
if c.displayed_by_default and not c.hidden and not c.never_hidden
|
234 |
+
],
|
235 |
+
label="Select columns to show",
|
236 |
+
elem_id="column-select",
|
237 |
+
interactive=True,
|
238 |
+
)
|
239 |
+
with gr.Row():
|
240 |
+
deleted_models_visibility = gr.Checkbox(
|
241 |
+
value=False, label="Show private/deleted models", interactive=True
|
242 |
+
)
|
243 |
+
merged_models_visibility = gr.Checkbox(
|
244 |
+
value=False, label="Show merges", interactive=True
|
245 |
+
)
|
246 |
+
flagged_models_visibility = gr.Checkbox(
|
247 |
+
value=False, label="Show flagged models", interactive=True
|
248 |
+
)
|
249 |
+
with gr.Column(min_width=320):
|
250 |
+
#with gr.Box(elem_id="box-filter"):
|
251 |
+
filter_columns_type = gr.CheckboxGroup(
|
252 |
+
label="Model types",
|
253 |
+
choices=[t.to_str() for t in ModelType],
|
254 |
+
value=[t.to_str() for t in ModelType],
|
255 |
+
interactive=True,
|
256 |
+
elem_id="filter-columns-type",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
257 |
)
|
258 |
+
filter_columns_precision = gr.CheckboxGroup(
|
259 |
+
label="Precision",
|
260 |
+
choices=[i.value.name for i in Precision],
|
261 |
+
value=[i.value.name for i in Precision],
|
262 |
+
interactive=True,
|
263 |
+
elem_id="filter-columns-precision",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
264 |
)
|
265 |
+
filter_columns_size = gr.CheckboxGroup(
|
266 |
+
label="Model sizes (in billions of parameters)",
|
267 |
+
choices=list(NUMERIC_INTERVALS.keys()),
|
268 |
+
value=list(NUMERIC_INTERVALS.keys()),
|
269 |
+
interactive=True,
|
270 |
+
elem_id="filter-columns-size",
|
|
|
271 |
)
|
272 |
+
filter_columns_add_special_tokens = gr.CheckboxGroup(
|
273 |
+
label="Add Special Tokens",
|
274 |
+
choices=[i.value.name for i in AddSpecialTokens],
|
275 |
+
value=[i.value.name for i in AddSpecialTokens],
|
276 |
+
interactive=True,
|
277 |
+
elem_id="filter-columns-add-special-tokens",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
278 |
)
|
279 |
+
filter_columns_num_few_shots = gr.CheckboxGroup(
|
280 |
+
label="Num Few Shots",
|
281 |
+
choices=[i.value.name for i in NumFewShots],
|
282 |
+
value=[i.value.name for i in NumFewShots],
|
283 |
+
interactive=True,
|
284 |
+
elem_id="filter-columns-num-few-shots",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
285 |
)
|
286 |
+
|
287 |
+
leaderboard_df_filtered = filter_models(leaderboard_df, [t.to_str(" : ") for t in ModelType], list(NUMERIC_INTERVALS.keys()), [i.value.name for i in Precision], [i.value.name for i in AddSpecialTokens], [i.value.name for i in NumFewShots], False, False, False)
|
288 |
+
# initial_columns = [c.name for c in fields(AutoEvalColumn) if c.never_hidden or c.displayed_by_default]
|
289 |
+
# leaderboard_df_filtered = select_columns(leaderboard_df_filtered, initial_columns)
|
290 |
+
|
291 |
+
# leaderboard_table = gr.components.Dataframe(
|
292 |
+
# value=leaderboard_df_filtered,
|
293 |
+
# headers=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + shown_columns.value,
|
294 |
+
# datatype=TYPES,
|
295 |
+
# elem_id="leaderboard-table",
|
296 |
+
# interactive=False,
|
297 |
+
# visible=True,
|
298 |
+
# )
|
299 |
+
|
300 |
+
# DataFrameの初期化部分のみを修正
|
301 |
+
initial_columns = ['T'] + [c.name for c in fields(AutoEvalColumn) if (c.never_hidden or c.displayed_by_default) and c.name != 'T']
|
302 |
+
leaderboard_df_filtered = select_columns(leaderboard_df, initial_columns)
|
303 |
+
|
304 |
+
# Model列のリンク形式を修正
|
305 |
+
leaderboard_df_filtered['Model'] = leaderboard_df_filtered['Model'].apply(
|
306 |
+
lambda x: f'[{x.split(">")[-2].split("<")[0]}]({x.split("href=")[1].split(chr(34))[1]})' if isinstance(x, str) and 'href=' in x else x
|
307 |
+
)
|
308 |
+
|
309 |
+
# 数値データを文字列に変換
|
310 |
+
for col in leaderboard_df_filtered.columns:
|
311 |
+
if col not in ['T', 'Model']:
|
312 |
+
leaderboard_df_filtered[col] = leaderboard_df_filtered[col].astype(str)
|
313 |
+
|
314 |
+
# DataFrameコンポーネントの初期化
|
315 |
+
leaderboard_table = gr.components.Dataframe(
|
316 |
+
value=leaderboard_df_filtered,
|
317 |
+
headers=initial_columns,
|
318 |
+
datatype=TYPES,
|
319 |
+
elem_id="leaderboard-table",
|
320 |
+
interactive=False,
|
321 |
+
visible=True
|
322 |
+
)
|
323 |
+
|
324 |
+
# Dummy leaderboard for handling the case when the user uses backspace key
|
325 |
+
hidden_leaderboard_table_for_search = gr.components.Dataframe(
|
326 |
+
value=original_df[COLS],
|
327 |
+
headers=COLS,
|
328 |
+
datatype=TYPES,
|
329 |
+
visible=False,
|
330 |
+
)
|
331 |
+
search_bar.submit(
|
332 |
+
update_table,
|
333 |
+
[
|
334 |
+
hidden_leaderboard_table_for_search,
|
335 |
+
shown_columns,
|
336 |
+
filter_columns_type,
|
337 |
+
filter_columns_precision,
|
338 |
+
filter_columns_size,
|
339 |
+
filter_columns_add_special_tokens,
|
340 |
+
filter_columns_num_few_shots,
|
341 |
+
deleted_models_visibility,
|
342 |
+
merged_models_visibility,
|
343 |
+
flagged_models_visibility,
|
344 |
+
search_bar,
|
345 |
+
],
|
346 |
+
leaderboard_table,
|
347 |
+
)
|
348 |
+
|
349 |
+
# Define a hidden component that will trigger a reload only if a query parameter has be set
|
350 |
+
hidden_search_bar = gr.Textbox(value="", visible=False)
|
351 |
+
hidden_search_bar.change(
|
352 |
+
update_table,
|
353 |
+
[
|
354 |
+
hidden_leaderboard_table_for_search,
|
355 |
+
shown_columns,
|
356 |
+
filter_columns_type,
|
357 |
+
filter_columns_precision,
|
358 |
+
filter_columns_size,
|
359 |
+
filter_columns_add_special_tokens,
|
360 |
+
filter_columns_num_few_shots,
|
361 |
+
deleted_models_visibility,
|
362 |
+
merged_models_visibility,
|
363 |
+
flagged_models_visibility,
|
364 |
+
search_bar,
|
365 |
+
],
|
366 |
+
leaderboard_table,
|
367 |
+
)
|
368 |
+
# Check query parameter once at startup and update search bar + hidden component
|
369 |
+
demo_leaderboard.load(load_query, inputs=[], outputs=[search_bar, hidden_search_bar])
|
370 |
+
|
371 |
+
for selector in [shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, filter_columns_add_special_tokens, filter_columns_num_few_shots, deleted_models_visibility, merged_models_visibility, flagged_models_visibility]:
|
372 |
+
selector.change(
|
373 |
+
update_table,
|
374 |
+
[
|
375 |
+
hidden_leaderboard_table_for_search,
|
376 |
+
shown_columns,
|
377 |
+
filter_columns_type,
|
378 |
+
filter_columns_precision,
|
379 |
+
filter_columns_size,
|
380 |
+
filter_columns_add_special_tokens,
|
381 |
+
filter_columns_num_few_shots,
|
382 |
+
deleted_models_visibility,
|
383 |
+
merged_models_visibility,
|
384 |
+
flagged_models_visibility,
|
385 |
+
search_bar,
|
386 |
+
],
|
387 |
+
leaderboard_table,
|
388 |
+
queue=True,
|
389 |
+
)
|
390 |
+
|
391 |
+
|
392 |
+
with gr.Blocks(css=custom_css) as demo:
|
393 |
+
gr.HTML(TITLE)
|
394 |
+
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
|
395 |
+
|
396 |
+
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
397 |
+
with gr.TabItem("🏅 LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
|
398 |
+
demo_leaderboard.render()
|
399 |
|
400 |
with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=2):
|
401 |
gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
|