Spaces:
Running
Running
natolambert
commited on
Commit
•
4a1518a
1
Parent(s):
fe666e1
add colors
Browse files
app.py
CHANGED
@@ -193,7 +193,36 @@ def random_sample(r: gr.Request, subset):
|
|
193 |
|
194 |
subsets = eval_set.unique("subset")
|
195 |
|
196 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
197 |
"""
|
198 |
Takes a model name as a regex, then returns only the rows that has that in it.
|
199 |
"""
|
@@ -228,6 +257,8 @@ def regex_table(dataframe, regex, filter_button):
|
|
228 |
# sort array by Score column
|
229 |
data = data.sort_values(by='Score', ascending=False)
|
230 |
|
|
|
|
|
231 |
# replace column '' with count/rank
|
232 |
data[''] = np.arange(1, 1 + len(data))
|
233 |
|
@@ -242,11 +273,15 @@ def regex_table(dataframe, regex, filter_button):
|
|
242 |
# replace any data[col].values == '' with np.NaN
|
243 |
data[col] = data[col].replace('', np.NaN)
|
244 |
data[col] = np.round(np.array(data[col].values).astype(float), 1)
|
|
|
|
|
|
|
|
|
245 |
return data
|
246 |
|
247 |
# import ipdb; ipdb.set_trace()
|
248 |
|
249 |
-
total_models = len(regex_table(rewardbench_data_avg.copy(), "", ["Seq. Classifiers", "DPO", "Custom Classifiers", "Generative"]).values)
|
250 |
|
251 |
with gr.Blocks(css=custom_css) as app:
|
252 |
# create tabs for the app, moving the current table to one titled "rewardbench" and the benchmark_text to a tab called "About"
|
@@ -281,7 +316,7 @@ with gr.Blocks(css=custom_css) as app:
|
|
281 |
visible=False,
|
282 |
)
|
283 |
rewardbench_table = gr.Dataframe(
|
284 |
-
regex_table(rewardbench_data_avg.copy(), "", ["Seq. Classifiers", "DPO", "Custom Classifiers", "Generative", "Prior Sets"])
|
285 |
datatype=col_types_rewardbench_avg,
|
286 |
headers=rewardbench_data_avg.columns.tolist(),
|
287 |
elem_id="rewardbench_dataframe_avg",
|
@@ -306,7 +341,7 @@ with gr.Blocks(css=custom_css) as app:
|
|
306 |
visible=False,
|
307 |
)
|
308 |
rewardbench_table_detailed = gr.Dataframe(
|
309 |
-
regex_table(rewardbench_data.copy(), "", ["Seq. Classifiers", "DPO", "Generative", "Custom Classifiers"])
|
310 |
datatype=col_types_rewardbench,
|
311 |
headers=rewardbench_data.columns.tolist(),
|
312 |
elem_id="rewardbench_dataframe",
|
@@ -351,7 +386,7 @@ with gr.Blocks(css=custom_css) as app:
|
|
351 |
visible=False,
|
352 |
)
|
353 |
pref_sets_table = gr.Dataframe(
|
354 |
-
regex_table(prefs_data.copy(), "", ["Seq. Classifiers", "DPO", "Custom Classifiers"])
|
355 |
datatype=col_types_prefs,
|
356 |
headers=prefs_data.columns.tolist(),
|
357 |
elem_id="prefs_dataframe",
|
|
|
193 |
|
194 |
subsets = eval_set.unique("subset")
|
195 |
|
196 |
+
color_map = {
|
197 |
+
"Generative": "#7497db",
|
198 |
+
"Custom Classifier": "#E8ECF2",
|
199 |
+
"Seq. Classifier": "#ffcd75",
|
200 |
+
"DPO": "#75809c",
|
201 |
+
}
|
202 |
+
def color_model_type_column(df, color_map):
|
203 |
+
"""
|
204 |
+
Apply color to the 'Model Type' column of the DataFrame based on a given color mapping.
|
205 |
+
|
206 |
+
Parameters:
|
207 |
+
df (pd.DataFrame): The DataFrame containing the 'Model Type' column.
|
208 |
+
color_map (dict): A dictionary mapping model types to colors.
|
209 |
+
|
210 |
+
Returns:
|
211 |
+
pd.Styler: The styled DataFrame.
|
212 |
+
"""
|
213 |
+
# Function to apply color based on the model type
|
214 |
+
def apply_color(val):
|
215 |
+
color = color_map.get(val, "default") # Default color if not specified in color_map
|
216 |
+
return f'background-color: {color}'
|
217 |
+
|
218 |
+
# Format for different columns
|
219 |
+
format_dict = {col: "{:.1f}" for col in df.columns if col not in ['Average', 'Model', 'Model Type']}
|
220 |
+
format_dict['Average'] = "{:.2f}"
|
221 |
+
format_dict[''] = "{:d}"
|
222 |
+
|
223 |
+
return df.style.applymap(apply_color, subset=['Model Type']).format(format_dict, na_rep='')
|
224 |
+
|
225 |
+
def regex_table(dataframe, regex, filter_button, style=True):
|
226 |
"""
|
227 |
Takes a model name as a regex, then returns only the rows that has that in it.
|
228 |
"""
|
|
|
257 |
# sort array by Score column
|
258 |
data = data.sort_values(by='Score', ascending=False)
|
259 |
|
260 |
+
data.reset_index(drop=True, inplace=True)
|
261 |
+
|
262 |
# replace column '' with count/rank
|
263 |
data[''] = np.arange(1, 1 + len(data))
|
264 |
|
|
|
273 |
# replace any data[col].values == '' with np.NaN
|
274 |
data[col] = data[col].replace('', np.NaN)
|
275 |
data[col] = np.round(np.array(data[col].values).astype(float), 1)
|
276 |
+
if style:
|
277 |
+
# apply color
|
278 |
+
data = color_model_type_column(data, color_map)
|
279 |
+
|
280 |
return data
|
281 |
|
282 |
# import ipdb; ipdb.set_trace()
|
283 |
|
284 |
+
total_models = len(regex_table(rewardbench_data_avg.copy(), "", ["Seq. Classifiers", "DPO", "Custom Classifiers", "Generative"], style=False).values)
|
285 |
|
286 |
with gr.Blocks(css=custom_css) as app:
|
287 |
# create tabs for the app, moving the current table to one titled "rewardbench" and the benchmark_text to a tab called "About"
|
|
|
316 |
visible=False,
|
317 |
)
|
318 |
rewardbench_table = gr.Dataframe(
|
319 |
+
regex_table(rewardbench_data_avg.copy(), "", ["Seq. Classifiers", "DPO", "Custom Classifiers", "Generative", "Prior Sets"]),
|
320 |
datatype=col_types_rewardbench_avg,
|
321 |
headers=rewardbench_data_avg.columns.tolist(),
|
322 |
elem_id="rewardbench_dataframe_avg",
|
|
|
341 |
visible=False,
|
342 |
)
|
343 |
rewardbench_table_detailed = gr.Dataframe(
|
344 |
+
regex_table(rewardbench_data.copy(), "", ["Seq. Classifiers", "DPO", "Generative", "Custom Classifiers"]),
|
345 |
datatype=col_types_rewardbench,
|
346 |
headers=rewardbench_data.columns.tolist(),
|
347 |
elem_id="rewardbench_dataframe",
|
|
|
386 |
visible=False,
|
387 |
)
|
388 |
pref_sets_table = gr.Dataframe(
|
389 |
+
regex_table(prefs_data.copy(), "", ["Seq. Classifiers", "DPO", "Custom Classifiers"]),
|
390 |
datatype=col_types_prefs,
|
391 |
headers=prefs_data.columns.tolist(),
|
392 |
elem_id="prefs_dataframe",
|