Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -41,9 +41,9 @@ def get_plots(task):
|
|
41 |
df = pd.read_csv('data/energy/' + task)
|
42 |
if df.columns[0].startswith("Unnamed:"):
|
43 |
df = df.iloc[:, 1:]
|
44 |
-
#
|
45 |
df['total_gpu_energy'] = df['total_gpu_energy'].astype(float)
|
46 |
-
# Convert energy_score to
|
47 |
df['energy_score'] = df['energy_score'].astype(int).astype(str)
|
48 |
df['Display Model'] = df['model'].apply(lambda m: m.split('/')[-1])
|
49 |
|
@@ -59,10 +59,11 @@ def get_plots(task):
|
|
59 |
width=800,
|
60 |
color_discrete_map=color_map
|
61 |
)
|
|
|
62 |
fig.update_traces(
|
63 |
hovertemplate="<br>".join([
|
64 |
"Model: %{y}",
|
65 |
-
"GPU Energy (Wh): %{x}",
|
66 |
"Energy Score: %{customdata[0]}"
|
67 |
])
|
68 |
)
|
@@ -96,7 +97,7 @@ def get_all_plots():
|
|
96 |
fig.update_traces(
|
97 |
hovertemplate="<br>".join([
|
98 |
"Model: %{y}",
|
99 |
-
"GPU Energy (Wh): %{x}",
|
100 |
"Energy Score: %{customdata[0]}"
|
101 |
])
|
102 |
)
|
@@ -108,16 +109,12 @@ def get_model_names(task):
|
|
108 |
if df.columns[0].startswith("Unnamed:"):
|
109 |
df = df.iloc[:, 1:]
|
110 |
df['energy_score'] = df['energy_score'].astype(int)
|
|
|
111 |
df['GPU Energy (Wh)'] = df['total_gpu_energy'].astype(float).apply(lambda x: f"{x:.4f}")
|
112 |
df['Model'] = df['model'].apply(make_link)
|
113 |
df['Score'] = df['energy_score'].apply(format_stars)
|
114 |
-
|
115 |
-
|
116 |
-
df['Class'] = df['class']
|
117 |
-
df = df[['Model', 'GPU Energy (Wh)', 'Score', 'Class']]
|
118 |
-
else:
|
119 |
-
df = df[['Model', 'GPU Energy (Wh)', 'Score']]
|
120 |
-
|
121 |
df = df.sort_values(by='GPU Energy (Wh)')
|
122 |
return df
|
123 |
|
@@ -139,7 +136,7 @@ def get_text_generation_plots(model_class):
|
|
139 |
df = pd.read_csv('data/energy/text_generation.csv')
|
140 |
if df.columns[0].startswith("Unnamed:"):
|
141 |
df = df.iloc[:, 1:]
|
142 |
-
# Filter
|
143 |
if 'class' in df.columns:
|
144 |
df = df[df['class'] == model_class]
|
145 |
df['total_gpu_energy'] = df['total_gpu_energy'].astype(float)
|
@@ -161,7 +158,7 @@ def get_text_generation_plots(model_class):
|
|
161 |
fig.update_traces(
|
162 |
hovertemplate="<br>".join([
|
163 |
"Model: %{y}",
|
164 |
-
"GPU Energy (Wh): %{x}",
|
165 |
"Energy Score: %{customdata[0]}"
|
166 |
])
|
167 |
)
|
@@ -178,11 +175,8 @@ def get_text_generation_model_names(model_class):
|
|
178 |
df['GPU Energy (Wh)'] = df['total_gpu_energy'].astype(float).apply(lambda x: f"{x:.4f}")
|
179 |
df['Model'] = df['model'].apply(make_link)
|
180 |
df['Score'] = df['energy_score'].apply(format_stars)
|
181 |
-
if
|
182 |
-
|
183 |
-
df = df[['Model', 'GPU Energy (Wh)', 'Score', 'Class']]
|
184 |
-
else:
|
185 |
-
df = df[['Model', 'GPU Energy (Wh)', 'Score']]
|
186 |
df = df.sort_values(by='GPU Energy (Wh)')
|
187 |
return df
|
188 |
|
@@ -218,14 +212,16 @@ Click through the tasks below to see how different models measure up in terms of
|
|
218 |
with gr.Tabs():
|
219 |
# --- Text Generation Tab with Dropdown for Model Class ---
|
220 |
with gr.TabItem("Text Generation 💬"):
|
|
|
|
|
|
|
|
|
221 |
with gr.Row():
|
222 |
with gr.Column(scale=1.3):
|
223 |
tg_plot = gr.Plot(get_text_generation_plots("A"))
|
224 |
with gr.Column(scale=1):
|
225 |
tg_table = gr.Dataframe(get_text_generation_model_names("A"), datatype="markdown")
|
226 |
-
|
227 |
-
label="Select Model Class",
|
228 |
-
value="A")
|
229 |
model_class_dropdown.change(fn=update_text_generation,
|
230 |
inputs=model_class_dropdown,
|
231 |
outputs=[tg_plot, tg_table])
|
|
|
41 |
df = pd.read_csv('data/energy/' + task)
|
42 |
if df.columns[0].startswith("Unnamed:"):
|
43 |
df = df.iloc[:, 1:]
|
44 |
+
# Convert GPU energy to float so that very small numbers are preserved
|
45 |
df['total_gpu_energy'] = df['total_gpu_energy'].astype(float)
|
46 |
+
# Convert energy_score to categorical string for proper discrete coloring
|
47 |
df['energy_score'] = df['energy_score'].astype(int).astype(str)
|
48 |
df['Display Model'] = df['model'].apply(lambda m: m.split('/')[-1])
|
49 |
|
|
|
59 |
width=800,
|
60 |
color_discrete_map=color_map
|
61 |
)
|
62 |
+
# GPU Energy now shows 4 decimals in the hover text
|
63 |
fig.update_traces(
|
64 |
hovertemplate="<br>".join([
|
65 |
"Model: %{y}",
|
66 |
+
"GPU Energy (Wh): %{x:.4f}",
|
67 |
"Energy Score: %{customdata[0]}"
|
68 |
])
|
69 |
)
|
|
|
97 |
fig.update_traces(
|
98 |
hovertemplate="<br>".join([
|
99 |
"Model: %{y}",
|
100 |
+
"GPU Energy (Wh): %{x:.4f}",
|
101 |
"Energy Score: %{customdata[0]}"
|
102 |
])
|
103 |
)
|
|
|
109 |
if df.columns[0].startswith("Unnamed:"):
|
110 |
df = df.iloc[:, 1:]
|
111 |
df['energy_score'] = df['energy_score'].astype(int)
|
112 |
+
# Ensure GPU Energy is a float and format it to 4 decimals
|
113 |
df['GPU Energy (Wh)'] = df['total_gpu_energy'].astype(float).apply(lambda x: f"{x:.4f}")
|
114 |
df['Model'] = df['model'].apply(make_link)
|
115 |
df['Score'] = df['energy_score'].apply(format_stars)
|
116 |
+
# Remove any Class column
|
117 |
+
df = df[['Model', 'GPU Energy (Wh)', 'Score']]
|
|
|
|
|
|
|
|
|
|
|
118 |
df = df.sort_values(by='GPU Energy (Wh)')
|
119 |
return df
|
120 |
|
|
|
136 |
df = pd.read_csv('data/energy/text_generation.csv')
|
137 |
if df.columns[0].startswith("Unnamed:"):
|
138 |
df = df.iloc[:, 1:]
|
139 |
+
# Filter by the selected model class if the "class" column exists
|
140 |
if 'class' in df.columns:
|
141 |
df = df[df['class'] == model_class]
|
142 |
df['total_gpu_energy'] = df['total_gpu_energy'].astype(float)
|
|
|
158 |
fig.update_traces(
|
159 |
hovertemplate="<br>".join([
|
160 |
"Model: %{y}",
|
161 |
+
"GPU Energy (Wh): %{x:.4f}",
|
162 |
"Energy Score: %{customdata[0]}"
|
163 |
])
|
164 |
)
|
|
|
175 |
df['GPU Energy (Wh)'] = df['total_gpu_energy'].astype(float).apply(lambda x: f"{x:.4f}")
|
176 |
df['Model'] = df['model'].apply(make_link)
|
177 |
df['Score'] = df['energy_score'].apply(format_stars)
|
178 |
+
# Remove the Class column if it exists
|
179 |
+
df = df[['Model', 'GPU Energy (Wh)', 'Score']]
|
|
|
|
|
|
|
180 |
df = df.sort_values(by='GPU Energy (Wh)')
|
181 |
return df
|
182 |
|
|
|
212 |
with gr.Tabs():
|
213 |
# --- Text Generation Tab with Dropdown for Model Class ---
|
214 |
with gr.TabItem("Text Generation 💬"):
|
215 |
+
# Dropdown moved above the plot and leaderboard
|
216 |
+
model_class_dropdown = gr.Dropdown(choices=["A", "B", "C"],
|
217 |
+
label="Select Model Class",
|
218 |
+
value="A")
|
219 |
with gr.Row():
|
220 |
with gr.Column(scale=1.3):
|
221 |
tg_plot = gr.Plot(get_text_generation_plots("A"))
|
222 |
with gr.Column(scale=1):
|
223 |
tg_table = gr.Dataframe(get_text_generation_model_names("A"), datatype="markdown")
|
224 |
+
# Update plot and table when the dropdown value changes
|
|
|
|
|
225 |
model_class_dropdown.change(fn=update_text_generation,
|
226 |
inputs=model_class_dropdown,
|
227 |
outputs=[tg_plot, tg_table])
|