ClΓ©mentine
commited on
Commit
β’
ecef2dc
1
Parent(s):
613696b
updated design to select columns to display
Browse files- app.py +28 -38
- src/utils_display.py +1 -1
app.py
CHANGED
@@ -259,16 +259,22 @@ def refresh():
|
|
259 |
)
|
260 |
|
261 |
|
262 |
-
def search_table(df, query):
|
263 |
-
if AutoEvalColumn.model_type.name in
|
264 |
filtered_df = df[
|
265 |
(df[AutoEvalColumn.dummy.name].str.contains(query, case=False))
|
266 |
| (df[AutoEvalColumn.model_type.name].str.contains(query, case=False))
|
267 |
]
|
268 |
else:
|
269 |
filtered_df = df[(df[AutoEvalColumn.dummy.name].str.contains(query, case=False))]
|
270 |
-
return filtered_df
|
|
|
271 |
|
|
|
|
|
|
|
|
|
|
|
272 |
|
273 |
def change_tab(query_param):
|
274 |
query_param = query_param.replace("'", '"')
|
@@ -288,44 +294,30 @@ demo = gr.Blocks(css=custom_css)
|
|
288 |
with demo:
|
289 |
gr.HTML(TITLE)
|
290 |
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
|
291 |
-
with gr.Row():
|
292 |
-
with gr.Box(elem_id="search-bar-table-box"):
|
293 |
-
search_bar = gr.Textbox(
|
294 |
-
placeholder="π Search your model and press ENTER...",
|
295 |
-
show_label=False,
|
296 |
-
elem_id="search-bar",
|
297 |
-
)
|
298 |
|
299 |
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
300 |
with gr.TabItem("π
LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
|
301 |
-
|
302 |
-
|
303 |
-
|
304 |
-
|
305 |
-
|
306 |
-
|
307 |
-
|
308 |
-
|
309 |
-
|
310 |
-
|
311 |
-
|
312 |
-
|
313 |
-
|
314 |
-
visible=False,
|
315 |
-
)
|
316 |
-
search_bar.submit(
|
317 |
-
search_table,
|
318 |
-
[hidden_leaderboard_table_for_search_lite, search_bar],
|
319 |
-
leaderboard_table_lite,
|
320 |
-
)
|
321 |
-
|
322 |
-
with gr.TabItem("π Extended model view", elem_id="llm-benchmark-tab-table", id=1):
|
323 |
leaderboard_table = gr.components.Dataframe(
|
324 |
-
value=leaderboard_df,
|
325 |
-
headers=
|
326 |
datatype=TYPES,
|
327 |
max_rows=None,
|
328 |
elem_id="leaderboard-table",
|
|
|
|
|
329 |
)
|
330 |
|
331 |
# Dummy leaderboard for handling the case when the user uses backspace key
|
@@ -338,9 +330,10 @@ with demo:
|
|
338 |
)
|
339 |
search_bar.submit(
|
340 |
search_table,
|
341 |
-
[hidden_leaderboard_table_for_search, search_bar],
|
342 |
leaderboard_table,
|
343 |
)
|
|
|
344 |
with gr.TabItem("π About", elem_id="llm-benchmark-tab-table", id=2):
|
345 |
gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
|
346 |
|
@@ -392,7 +385,6 @@ with demo:
|
|
392 |
label="Model type",
|
393 |
multiselect=False,
|
394 |
value="pretrained",
|
395 |
-
max_choices=1,
|
396 |
interactive=True,
|
397 |
)
|
398 |
|
@@ -402,7 +394,6 @@ with demo:
|
|
402 |
label="Precision",
|
403 |
multiselect=False,
|
404 |
value="float16",
|
405 |
-
max_choices=1,
|
406 |
interactive=True,
|
407 |
)
|
408 |
weight_type = gr.Dropdown(
|
@@ -410,7 +401,6 @@ with demo:
|
|
410 |
label="Weights type",
|
411 |
multiselect=False,
|
412 |
value="Original",
|
413 |
-
max_choices=1,
|
414 |
interactive=True,
|
415 |
)
|
416 |
base_model_name_textbox = gr.Textbox(
|
|
|
259 |
)
|
260 |
|
261 |
|
262 |
+
def search_table(df, leaderboard_table, query):
|
263 |
+
if AutoEvalColumn.model_type.name in leaderboard_table.columns:
|
264 |
filtered_df = df[
|
265 |
(df[AutoEvalColumn.dummy.name].str.contains(query, case=False))
|
266 |
| (df[AutoEvalColumn.model_type.name].str.contains(query, case=False))
|
267 |
]
|
268 |
else:
|
269 |
filtered_df = df[(df[AutoEvalColumn.dummy.name].str.contains(query, case=False))]
|
270 |
+
return filtered_df[leaderboard_table.columns]
|
271 |
+
|
272 |
|
273 |
+
def select_columns(df, columns):
|
274 |
+
always_here_cols = [AutoEvalColumn.model_type_symbol.name, AutoEvalColumn.model.name]
|
275 |
+
# We use COLS to maintain sorting
|
276 |
+
filtered_df = df[always_here_cols + [c for c in COLS if c in df.columns and c in columns] + [AutoEvalColumn.dummy.name]]
|
277 |
+
return filtered_df
|
278 |
|
279 |
def change_tab(query_param):
|
280 |
query_param = query_param.replace("'", '"')
|
|
|
294 |
with demo:
|
295 |
gr.HTML(TITLE)
|
296 |
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
297 |
|
298 |
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
299 |
with gr.TabItem("π
LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
|
300 |
+
with gr.Row():
|
301 |
+
shown_columns = gr.CheckboxGroup(
|
302 |
+
choices = [c for c in COLS if c not in [AutoEvalColumn.dummy.name, AutoEvalColumn.model.name, AutoEvalColumn.model_type_symbol.name]],
|
303 |
+
value = [c for c in COLS_LITE if c not in [AutoEvalColumn.dummy.name, AutoEvalColumn.model.name, AutoEvalColumn.model_type_symbol.name]],
|
304 |
+
label="Select columns to show",
|
305 |
+
elem_id="column-select",
|
306 |
+
interactive=True,
|
307 |
+
)
|
308 |
+
search_bar = gr.Textbox(
|
309 |
+
placeholder="π Search for your model and press ENTER...",
|
310 |
+
show_label=False,
|
311 |
+
elem_id="search-bar",
|
312 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
313 |
leaderboard_table = gr.components.Dataframe(
|
314 |
+
value=leaderboard_df[[AutoEvalColumn.model_type_symbol.name, AutoEvalColumn.model.name] + shown_columns.value+ [AutoEvalColumn.dummy.name]],
|
315 |
+
headers=[AutoEvalColumn.model_type_symbol.name, AutoEvalColumn.model.name] + shown_columns.value + [AutoEvalColumn.dummy.name],
|
316 |
datatype=TYPES,
|
317 |
max_rows=None,
|
318 |
elem_id="leaderboard-table",
|
319 |
+
interactive=False,
|
320 |
+
visible=True,
|
321 |
)
|
322 |
|
323 |
# Dummy leaderboard for handling the case when the user uses backspace key
|
|
|
330 |
)
|
331 |
search_bar.submit(
|
332 |
search_table,
|
333 |
+
[hidden_leaderboard_table_for_search, leaderboard_table, search_bar],
|
334 |
leaderboard_table,
|
335 |
)
|
336 |
+
shown_columns.change(select_columns, [hidden_leaderboard_table_for_search, shown_columns], leaderboard_table)
|
337 |
with gr.TabItem("π About", elem_id="llm-benchmark-tab-table", id=2):
|
338 |
gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
|
339 |
|
|
|
385 |
label="Model type",
|
386 |
multiselect=False,
|
387 |
value="pretrained",
|
|
|
388 |
interactive=True,
|
389 |
)
|
390 |
|
|
|
394 |
label="Precision",
|
395 |
multiselect=False,
|
396 |
value="float16",
|
|
|
397 |
interactive=True,
|
398 |
)
|
399 |
weight_type = gr.Dropdown(
|
|
|
401 |
label="Weights type",
|
402 |
multiselect=False,
|
403 |
value="Original",
|
|
|
404 |
interactive=True,
|
405 |
)
|
406 |
base_model_name_textbox = gr.Textbox(
|
src/utils_display.py
CHANGED
@@ -20,7 +20,7 @@ class AutoEvalColumn: # Auto evals column
|
|
20 |
arc = ColumnContent("ARC", "number", True)
|
21 |
hellaswag = ColumnContent("HellaSwag", "number", True)
|
22 |
mmlu = ColumnContent("MMLU", "number", True)
|
23 |
-
truthfulqa = ColumnContent("TruthfulQA
|
24 |
model_type = ColumnContent("Type", "str", False)
|
25 |
precision = ColumnContent("Precision", "str", False, True)
|
26 |
license = ColumnContent("Hub License", "str", False)
|
|
|
20 |
arc = ColumnContent("ARC", "number", True)
|
21 |
hellaswag = ColumnContent("HellaSwag", "number", True)
|
22 |
mmlu = ColumnContent("MMLU", "number", True)
|
23 |
+
truthfulqa = ColumnContent("TruthfulQA", "number", True)
|
24 |
model_type = ColumnContent("Type", "str", False)
|
25 |
precision = ColumnContent("Precision", "str", False, True)
|
26 |
license = ColumnContent("Hub License", "str", False)
|