Spaces:
Running
Running
Jellyfish042
commited on
Merge remote-tracking branch 'origin/main'
Browse files- app.py +19 -13
- data/2024-07/14b.xlsx +0 -0
app.py
CHANGED
@@ -164,7 +164,8 @@ def update_table(period: str,
|
|
164 |
if len(combined_data) > 0:
|
165 |
sorted_data = combined_data.sort_values(by=sort_by, ascending=ascending)
|
166 |
sorted_data = sorted_data.rename(columns={'Average (The lower the better)': 'Average (lower=better)'})
|
167 |
-
|
|
|
168 |
filtered_data = sorted_data[visible_columns]
|
169 |
|
170 |
filtered_data.columns = [col.replace('_', ' ') for col in filtered_data.columns]
|
@@ -178,7 +179,7 @@ def update_table(period: str,
|
|
178 |
vmin = {}
|
179 |
vmax = {}
|
180 |
for column in filtered_data.columns:
|
181 |
-
if column in ['Name', '
|
182 |
continue
|
183 |
col_values = filtered_data[column]
|
184 |
if len(col_values) > 1:
|
@@ -191,9 +192,11 @@ def update_table(period: str,
|
|
191 |
target_color_columns.append('Average (lower=better)')
|
192 |
if 'Individual Tests' in color_columns:
|
193 |
target_color_columns.extend([col for col in filtered_data.columns if
|
194 |
-
col not in ['Name', '
|
|
|
|
|
|
|
195 |
|
196 |
-
styler = filtered_data.style.format(formatter)
|
197 |
for column in target_color_columns:
|
198 |
if column in vmin and column in vmax: # Ensure that the vmin and vmax dicts contain the column
|
199 |
styler = styler.background_gradient(cmap=cmap, subset=[column], vmin=vmin[column], vmax=vmax[column])
|
@@ -271,7 +274,8 @@ def create_scaling_plot(all_data, period):
|
|
271 |
names_to_connect = ['Meta-Llama-3-8B',
|
272 |
'stablelm-3b-4e1t',
|
273 |
'Qwen2-1.5B',
|
274 |
-
'TinyLlama-1.1B-intermediate-step-1431k-3T'
|
|
|
275 |
connection_points = new_df[new_df['Name'].isin(names_to_connect)]
|
276 |
|
277 |
new_df['Color'] = new_df['Name'].apply(lambda name: '#39C5BB' if name in names_to_connect else '#636efa')
|
@@ -284,7 +288,7 @@ def create_scaling_plot(all_data, period):
|
|
284 |
|
285 |
x_min = connection_points['Log Params(B)'].min()
|
286 |
x_max = connection_points['Log Params(B)'].max()
|
287 |
-
extended_x = np.linspace(x_min, x_max * 1.
|
288 |
extended_x_original = np.exp(extended_x)
|
289 |
trend_line_y = model.predict(extended_x.reshape(-1, 1))
|
290 |
trend_line_y_original = np.exp(trend_line_y)
|
@@ -347,8 +351,11 @@ def read_all_data(folder_name):
|
|
347 |
|
348 |
all_data, time_list = read_all_data('data')
|
349 |
|
350 |
-
|
351 |
-
|
|
|
|
|
|
|
352 |
initial_models = model_size_list
|
353 |
initial_metric = metric_list[0]
|
354 |
initial_columns = get_unique_column_names(all_data)
|
@@ -379,7 +386,7 @@ with gr.Blocks(css=css) as demo:
|
|
379 |
with gr.Tab("π Leaderboard"):
|
380 |
with gr.Row():
|
381 |
with gr.Column():
|
382 |
-
period_selector = gr.Dropdown(label="Period", choices=time_list, value=
|
383 |
model_selector = gr.CheckboxGroup(label="Model", choices=model_size_list, value=model_size_list)
|
384 |
metric_selector = gr.Dropdown(label="Metric", choices=metric_list, value=metric_list[0])
|
385 |
with gr.Column():
|
@@ -390,7 +397,7 @@ with gr.Blocks(css=css) as demo:
|
|
390 |
choices=get_unique_column_names(all_data),
|
391 |
value=get_unique_column_names(all_data))
|
392 |
|
393 |
-
table = gr.Dataframe(initial_data, column_widths=[130,
|
394 |
wrap=True,
|
395 |
height=800,
|
396 |
)
|
@@ -414,14 +421,13 @@ with gr.Blocks(css=css) as demo:
|
|
414 |
with gr.Tab("π MultiLang"):
|
415 |
gr.Markdown("## Coming soon...")
|
416 |
with gr.Tab("π Scaling Law"):
|
417 |
-
|
418 |
-
|
419 |
|
420 |
def update_plot(period):
|
421 |
new_fig = create_scaling_plot(all_data, period)
|
422 |
return new_fig
|
423 |
|
424 |
-
|
425 |
plot = gr.Plot(initial_fig)
|
426 |
period_selector_2.change(update_plot, inputs=period_selector_2, outputs=plot)
|
427 |
|
|
|
164 |
if len(combined_data) > 0:
|
165 |
sorted_data = combined_data.sort_values(by=sort_by, ascending=ascending)
|
166 |
sorted_data = sorted_data.rename(columns={'Average (The lower the better)': 'Average (lower=better)'})
|
167 |
+
sorted_data = sorted_data.rename(columns={'Parameters Count (B)': 'Params (B)'})
|
168 |
+
visible_columns = ['Name', 'Params (B)', 'Average (lower=better)'] + visible_columns
|
169 |
filtered_data = sorted_data[visible_columns]
|
170 |
|
171 |
filtered_data.columns = [col.replace('_', ' ') for col in filtered_data.columns]
|
|
|
179 |
vmin = {}
|
180 |
vmax = {}
|
181 |
for column in filtered_data.columns:
|
182 |
+
if column in ['Name', 'Params (B)']:
|
183 |
continue
|
184 |
col_values = filtered_data[column]
|
185 |
if len(col_values) > 1:
|
|
|
192 |
target_color_columns.append('Average (lower=better)')
|
193 |
if 'Individual Tests' in color_columns:
|
194 |
target_color_columns.extend([col for col in filtered_data.columns if
|
195 |
+
col not in ['Name', 'Params (B)', 'Average (lower=better)']])
|
196 |
+
|
197 |
+
|
198 |
+
styler = filtered_data.style.format(formatter).applymap(color_cell, subset=['Params (B)'])
|
199 |
|
|
|
200 |
for column in target_color_columns:
|
201 |
if column in vmin and column in vmax: # Ensure that the vmin and vmax dicts contain the column
|
202 |
styler = styler.background_gradient(cmap=cmap, subset=[column], vmin=vmin[column], vmax=vmax[column])
|
|
|
274 |
names_to_connect = ['Meta-Llama-3-8B',
|
275 |
'stablelm-3b-4e1t',
|
276 |
'Qwen2-1.5B',
|
277 |
+
'TinyLlama-1.1B-intermediate-step-1431k-3T',
|
278 |
+
'Mistral-Nemo-Base-2407']
|
279 |
connection_points = new_df[new_df['Name'].isin(names_to_connect)]
|
280 |
|
281 |
new_df['Color'] = new_df['Name'].apply(lambda name: '#39C5BB' if name in names_to_connect else '#636efa')
|
|
|
288 |
|
289 |
x_min = connection_points['Log Params(B)'].min()
|
290 |
x_max = connection_points['Log Params(B)'].max()
|
291 |
+
extended_x = np.linspace(x_min, x_max * 1.5, 100)
|
292 |
extended_x_original = np.exp(extended_x)
|
293 |
trend_line_y = model.predict(extended_x.reshape(-1, 1))
|
294 |
trend_line_y_original = np.exp(trend_line_y)
|
|
|
351 |
|
352 |
all_data, time_list = read_all_data('data')
|
353 |
|
354 |
+
time_list.sort()
|
355 |
+
last_period = time_list[-1]
|
356 |
+
|
357 |
+
initial_fig = create_scaling_plot(all_data, last_period)
|
358 |
+
initial_period = last_period
|
359 |
initial_models = model_size_list
|
360 |
initial_metric = metric_list[0]
|
361 |
initial_columns = get_unique_column_names(all_data)
|
|
|
386 |
with gr.Tab("π Leaderboard"):
|
387 |
with gr.Row():
|
388 |
with gr.Column():
|
389 |
+
period_selector = gr.Dropdown(label="Period", choices=time_list, value=last_period)
|
390 |
model_selector = gr.CheckboxGroup(label="Model", choices=model_size_list, value=model_size_list)
|
391 |
metric_selector = gr.Dropdown(label="Metric", choices=metric_list, value=metric_list[0])
|
392 |
with gr.Column():
|
|
|
397 |
choices=get_unique_column_names(all_data),
|
398 |
value=get_unique_column_names(all_data))
|
399 |
|
400 |
+
table = gr.Dataframe(initial_data, column_widths=[130, 50, 50, 35, 35, 35, 35, 35, 35, 35],
|
401 |
wrap=True,
|
402 |
height=800,
|
403 |
)
|
|
|
421 |
with gr.Tab("π MultiLang"):
|
422 |
gr.Markdown("## Coming soon...")
|
423 |
with gr.Tab("π Scaling Law"):
|
424 |
+
print(time_list)
|
425 |
+
period_selector_2 = gr.Dropdown(label="Period", choices=time_list, value=last_period)
|
426 |
|
427 |
def update_plot(period):
|
428 |
new_fig = create_scaling_plot(all_data, period)
|
429 |
return new_fig
|
430 |
|
|
|
431 |
plot = gr.Plot(initial_fig)
|
432 |
period_selector_2.change(update_plot, inputs=period_selector_2, outputs=plot)
|
433 |
|
data/2024-07/14b.xlsx
CHANGED
Binary files a/data/2024-07/14b.xlsx and b/data/2024-07/14b.xlsx differ
|
|